Define a promotion, select multiple products, write the copy, and send it to your whole customer database – that’s building email campaigns the old way. Different customers receive the same message, even when their behavior, preferences, and intent are completely different, often within the same segment. 

What if, instead of sending the same version of a campaign to everyone, each customer received a version tailored to what they’ve browsed, bought, and are likely to do next?

Let us show you what’s possible when you change the approach with the help of AI.

Step 1: the campaign

In our example, let’s look at a global activewear brand with a strong eCommerce presence and a CRM system already in place. They run regular promotional email campaigns and have a solid setup:

• Customer data from online and in-store purchases
• Lifecycle-based segmentation
• Automated flows and campaigns.

However, most campaigns are still built using rules like segments based on past purchases or activity, predefined product selections, and fixed messaging per audience group, which creates a big personalization gap. 

Customers shopping for activewear (or any other product) can behave very differently and be looking for different things, yet they receive the same email. The data about their purchase history, preferences, engagement, and browsing and shopping behavior is there, but it isn’t being used to decide what each customer should see. 

Step 2: specific use case

The impact of personalizing emails is highest at specific moments in the customer lifecycle.

Here’s a common scenario:

A customer has signed up, browsed products, maybe even made an initial purchase, but hasn’t come back in a while. Purchase intent is uncertain, and churn risk is increasing.

Also read:
Case Study: Conversational Commerce Brings 31% of Churned Customers Back

Most CRM setups already account for this with a predefined flow: Identify inactive users (e.g., no purchase in the last 30 days) → Place them into a re-engagement segment → Send a reminder or offer.

What you need to take into account is that within this “at-risk” group, customers are very different. They are grouped by inactivity, but instead of treating them as a single audience, it becomes a starting point for understanding what each individual customer is missing and what would motivate them to return.

Step 3: customer behavior becomes a clear profile

To personalize email at an individual level, you first need to move beyond segments and look at actual behavior. For each customer, the goal is to build a simple, usable profile based on what they’ve already done.

The best part is, this doesn’t require new data. It uses what most brands already have – purchase history, browsing activity, product categories viewed or bought, engagement with past emails or campaigns, etc. 

From this, you can start to understand patterns, for example:

• A customer consistently browsing the studio and low-impact categories
• A customer buying matching sets and newer collections
• A customer focused on discounted or everyday items
• A customer who browses often but hasn’t purchased yet.

These patterns help answer a practical and significant question – what is this customer actually interested in right now? At this stage, you want to create a clear enough picture to guide the next decision, even if you don’t predict everything perfectly.

Instead of assigning someone to a broad segment, you’re defining what they prefer, how they shop, and where they are in their journey. This profile becomes the input for personalization and as the foundation for deciding what each customer should see next.

Step 4: making it actionable with AI

Once you have a clear view of customer behavior, the next step is to decide what makes the most sense for that specific customer to see. Here’s where AI becomes the most useful. 

Instead of relying on predefined rules (e.g., “if customer bought X, show Y”), AI can evaluate multiple signals at once and determine which product is most relevant, what type of message fits best, and how to position the offer.

For example:

• A customer focused on studio workouts → show soft, flexible leggings → message around comfort and movement
• A customer interested in styling and trends → show a matching set → message around versatility and looks
• A customer buying casual pieces → show a jacket or layer → message around everyday wear
• A customer who hasn’t purchased yet → show an entry product → message focused on ease and value.

You’re no longer selecting products and writing one message for a segment. You’re letting the system decide, per customer, what to highlight, why it matters, how to present it, and maximize engagement. Each customer gets a direction that reflects their own behavior, and raw data becomes something actionable.

Step 5: email generation for each customer

Once the decision is made (product + angle + message), the email itself becomes straightforward. The system generates a version of the email for each customer using the selected product, the chosen message angle, a subject line that matches the intent, and a customized CTA. The structure of the email remains consistent, but the product, tone, and reasons to buy are tailored to each individual.

• The studio-focused customer receives: 
  • soft leggings
  • subject line around comfort or movement
  • copy that reflects low-impact workouts

• The trend-driven customer receives:
  • a matching set
  • subject line focused on style
  • copy that leans into versatility and looks

• The casual shopper receives:
  • a jacket
  • subject line around everyday wear
  • copy positioned around ease and layering

At this point, the campaign functions as a system that automatically produces many variations:

Step 6: strategy implementation within CRM

To make this happen, you don’t need to replace your current tools. The CRM still triggers the campaign, defines the audience (e.g., at-risk users), builds emails, and handles delivery.

What changes is what happens before the email is sent. Instead of pulling fixed templates and predefined product blocks, the system generates personalized content with tailored recommendations.

So the setup becomes:

• CRM → defines when and to whom
• AI layer → decides what each person sees.

The data is already there, and the campaigns already exist. What changes are the decisions before the email is sent. When each customer sees products and messaging that match how they actually shop, you will notice the difference immediately in higher engagement, clicks, conversions, and revenue.

Curious how this would work for your business? Contact our AI & email marketing strategists today, and we’ll show you concrete examples based on your products and customer purchase patterns, and show what your campaigns could look like with AI-driven personalization.

The post How to Personalize Email Campaign Communication with AI appeared first on scandiweb.

The demand for ad creatives has never been higher, and neither has the pressure to produce them faster. Whether running 5 campaigns or 50 at your company, you’re probably asking a similar question to everyone else right now: 

How are we going to complete all these visuals on time?

Creative production has traditionally been a bottleneck (especially with Black Friday or any sales season approaching), requiring styled photoshoots, stock image hunting, endless design rounds, and so on. With generative AI, there’s now a better way to approach this.

At scandiweb, we’ve trained and tested various AI tools to create high-performing visuals for ads. In fact, we produced over 100 high-performing ad creatives for different clients in just two weeks. Let’s explore how to use AI to your advantage – we’ll showcase real examples, share performance insights, and give you a blueprint to speed up creative production.

Sneak peek:)

Key takeaways

• AI ad creatives can match and often outperform traditional production, without the need for photoshoots, product shoots, stock images, or lengthy design timelines.
• What separates effective AI ads from the noise is clarity of intent.
• AI ads benefit from rapid testing and iteration.
• In real campaigns, AI creatives delivered high CTRs and low CPCs, proving their value across industries and formats.

The problem: always waiting on visuals

As digital ad platforms change and competition intensifies, brands are expected to appear everywhere, all the time. Traditional creative production can’t keep up. Producing a single set of high-quality visuals typically involves:

• Product photography or styled shoots
• Creative briefings and revisions
• Copy and design iterations
• Format adjustments for multiple channels.

Multiply that by every target audience, product line, message variant, and platform format, and suddenly your one campaign needs dozens of ad assets just to launch. That can result in:

Delays in launching campaigns because assets aren’t ready
Missed testing opportunities, limiting performance insights
Creative fatigue, with reused or generic visuals
Lower ROAS, simply due to visual underperformance.

Your media spend is only as strong as the creatives behind it. Luckily, there’s now a faster, smarter way to deliver them.

The solution: AI-generated ad creatives

Integrating AI into the ad production workflow allows for keeping up with any campaign schedules. Using tools like MidJourney, Runway, Veo3, ChatGPT, and Flora, our Paid Media team has built entire campaign asset libraries from scratch. No product photos and styled shoots needed for any of our clients.

Lifestyle banners, background visuals for motion ads, demos, side-by-side comparisons, or new assets to test messaging variations – with the right approach, you can have full control over the output.

When to use AI for ad creatives

Use AI-generated ad creatives when:

• You need to launch quickly without waiting on styled shoots or internal design resources
• You want to test multiple messaging angles (value-based, product-first, social proof, etc.)
• You’re running seasonal or short-term campaigns
• You want creative diversity
• You’re scaling to various platforms and audiences.

AI is ideal for creating all kinds of ads, including lifestyle visuals that place the product in a real-life context, demo-style images and UI overlays, social proof formats (reviews, ratings, testimonials), background visuals and motion graphics, and conceptual or metaphorical visuals that draw attention.

That said, AI isn’t a fit for every use case. Highly regulated industries or products that require precise accuracy might still need human-led design. But for the majority of brands, proper use of AI offers an advantage that’s hard to ignore.

AI in your campaign workflow

Here’s how a modern, AI-powered creative workflow might look:

• Define the campaign goal – are you building awareness or converting repeat customers? This determines the creative type.
• Generate visual concepts with AI tools – use text-to-image prompts, product references, or overlays to create a variety of visuals. 
• Test multiple variants – launch with 2–3 creative directions across platforms. Focus on A/B testing hooks, imagery, and formats.
• Refine based on data – double down on what’s working, then re-generate and adjust.
• Once you find a winning format, replicate it.

The faster you create, the more you can refine and witness performance start to build. This process would otherwise take weeks, but it’s significantly faster when assisted by AI. It’s particularly powerful when paired with high-volume campaign calendars, such as Black Friday, back-to-school sales, spring launches, or other seasonal sales campaigns.

Take a look at some of the stand-outs from our AI Creative Library!

How to make AI creatives that work

You have to remember users are currently hit with AI-generated content from every direction. Ads that work won’t be flashy but thoughtful. Realism plus subtle creative flair will build trust where gimmicks won’t. Think cool transitions and unexpected metaphors to catch attention. 

AI can generate anything. That’s the problem and the opportunity. Before you start prompting, ask “what’s the ad supposed to do?” Build trust? Educate, inspire action, show the product in real life? Reassure a skeptical buyer? The goal dictates the style and format of the creative.

Here’s how to make sure your AI-powered ads actually convert. 

Focus on user-led storytelling

One of the most effective creative angles right now? Show, don’t tell.

• Use AI to simulate UGC-style visuals, product-in-use scenarios, or day-in-the-life settings
• Emphasize the real-world benefit, not just the product
• Add elements like quotes, testimonials, or messaging overlays that anchor the creative in relatability
• Even if the visuals are not real, they should feel grounded in everyday life.

Add a visual twist

AI excels at creating subtle surrealism, but use that power strategically to catch attention without confusing the viewer:

• A background that moves while the product remains still
• A visual metaphor that conveys a product benefit
• A juxtaposition of two ideas (e.g., natural ingredients + clinical effectiveness).

If something looks too fake or too AI, it won’t convert. Shoppers are savvy, so you should still prioritize realistic lighting and proportions, avoid overly stylized or uncanny imagery, use reference images for consistency and brand alignment, and blend AI visuals with human-written copy and proven ad formats.

A/B test fast and often

Because AI allows you to generate multiple versions in hours, not days, you can quickly test different product angles, color tones, background context, emotional vs. rational creative tone, etc. Use performance data to refine your prompts. 

AI can create high-quality, on-brand visuals tailored to the campaign goal super fast. What’s more, our testing proves they work. Let us show you!

Case studies: AI ad creatives in action

Trust-building creatives for pet products

AI model: ChatGPT image generation with product reference

Goal: Build trust and conversions through social proof, tailored for pet owners

• Impressions: 15,312
• Click-through rate: 0.63%
• Return on ad spend: 10.55

Why it worked: realistic, relatable imagery paired with a genuine-sounding customer review hits home with pet owners, especially those dealing with sensitive digestion issues.

AI model: Runway

Goal: Educate and convert by visualizing high-quality ingredients

• Impressions: 39,686
• CTR: 0.50%
• ROAS: 9.50

Why it worked: ingredient awareness is a key driver for health-conscious shoppers; the visuals convey freshness and quality without needing photography.

Credibility in ads for a wellness brand

AI model: Flora GPT, Sora, Midjourney

Goal: Encourage action by showcasing consultation with Swiss doctors

• Reach: 160,158
• CPC: 3.14 CHF
• CPM: 1.61 CHF

Why it worked: emphasizing credibility helps move users from awareness to action, while visual realism plus healthcare-centric messaging builds trust.

AI model: Flora GPT, Midjourney, Runway

Goal: Inspire confidence and conversions for MS symptom management services

• Reach: 202,395
• CPC: 3.20 CHF
• CPM: 1.67 CHF

Why it worked: a blend of hopeful messaging and empowering visuals, conveying discretion, ease-of-use, and professionalism.

Video content enriched with AI

Goal: Drive efficient clicks and form submissions at low cost

• CTR: 3.65%
• Link Clicks: 235
• Reach: 4,860

Why it worked: combining real footage with AI-enhanced visual storytelling created credibility while catching attention. 

Goal: Capture attention and drive contact signups through a storytelling-led format

• CTR: 3.45%
• Link Clicks: 207
• Reach: 4,786

Why it worked: the animated format created an emotional hook while staying brand-appropriate. It delivered a relatable, human feel, resulting in the highest contact volume of the test group.

So, it’s official, AI has entered the paid media toolkit as a high-performance solution. If properly trained, AI-generated visuals can be on-brand, performance-driven, and scalable across platforms, converting effectively.

We’ve seen firsthand how AI-generated ad creatives:

• Help teams produce campaign assets faster and at higher volume
• Deliver visuals that align with brand and messaging goals, without needing custom photos or stock images
• Enable faster testing cycles, improving creative insights and campaign outcomes
• Perform across different formats – motion graphics, lifestyle ads, review-based visuals, and so much more.

Ready to launch your best-performing ads with ease? Our AI and Paid Media teams have already created hundreds of high-performing creatives for clients across industries, and we’re just getting started. Book a free consultation, and let’s figure out how AI can improve your ads and your broader marketing operations!

The post These Ad Banners Are AI-Generated, but the Clicks Are Real appeared first on scandiweb.

The PPC team was spending over an hour conducting manual PPC audits and health checks. These tasks require working with large sets of pure data—time-consuming and repetitive work that doesn’t leverage human creativity.

To streamline this, we automated the process, letting AI handle the heavy lifting of data analysis while embedding best practices directly into the workflow.

How it works

We built a Claude project that connects to an n8n MCP server, which hosts a range of Google Ads tools. This setup allows Claude to execute PPC audits end-to-end with minimal manual effort.

The project’s knowledge base includes audit templates, best practices, and different audit types—helping Claude not only analyze but also present the results in a structured, accessible way.

Processing and development

Input

• Google Ads client account ID—thats it!

Output

• Comprehensive PPC Audit and breakdown
• Standardized audit format, reducing variability between team members
• Actionable recommendations aligned with industry best practices

Impact

What previously took over an hour of manual work can now be completed in minutes. The PPC team can scale audits, ensure consistency, and focus more on client welfare than pure data.

Built with

• Claude – for analysis and structured reporting
• n8n MCP server – hosting the audit toolset and handling orchestration
• Google Ads API – providing real-time campaign data

Complexity

Medium

Looking to leverage modern AI tools within your company? Get in touch and explore next steps.

The post AI Agent Bites #10: Automating PPC Audits with Claude and MCP appeared first on scandiweb.

There can be a variety of reasons why customers stop buying. The order gets delayed, a routine changes, life simply gets in the way, and a solid reminder never comes. Traditional win-back tactics like email campaigns or promo blasts often feel impersonal and easy to ignore. It’s easy to assume they’re gone for good.

While most teams focus on acquiring new customers, 65% of a company’s business typically comes from existing ones. And you lose revenue when that group goes inactive. 

Conversational commerce can make a difference. It offers a way to reach churned customers with real-time, personalized messages in their most used channels. With no manual outreach! Once set up, these messages are triggered and adapted automatically based on each customer’s data.

Our client, Zorro.lv, a growing pet food retailer, took this approach. Instead of scaling with bulk messaging, they chose to replicate the one-on-one conversations that defined their brand from the start.

About

Zorro.lv built its reputation on offering high‑quality, natural pet food and treats without fillers and artificial additives. Their products include grain‑free and super‑premium lines, tailored for pets with different nutritional needs. Zorro.lv aims to combine high‑quality products with an attentive, trustworthy customer experience.

As their customers multiplied, the team could no longer personally reach out to those who’d drifted away. We partnered with them to build an AI‑powered sales concierge that restored personal attention via WhatsApp.

Project goals

Zorro.lv had always relied on direct, personal outreach to build loyalty. Now, the team needed a way to:

• Identify the right customer segment most likely to purchase again
• Reconnect with lapsed customers in a personal and timely way
• Maintain Zorro’s relationship-driven tone without switching to bulk promotions
• Automate outreach at scale without losing authenticity
• Free up sales and support staff from repetitive follow-ups while keeping the customer experience consistent.

The ultimate goal was to recreate the feeling of a dedicated salesperson for every customer. Someone who knows their order history, understands their pet’s needs, and reaches out at the right moment.

Approach

To bring back churned customers without losing Zorro.lv’s personal touch, we developed an AI-powered sales concierge connected to their Shopify store and WhatsApp. The project began with building a clear customer segment, then moved into conversation design and technical integration to make outreach feel natural for every customer.

Customer selection

We built the first outreach segment from Zorro.lv’s Shopify data and ensured outreach went to high-potential customers who already had a relationship with the brand. Customers were included if they:

• Placed their last order between 12–24 months ago
• Spent at least €100 in total
• Completed at least two purchases
• Gave marketing consent.

Conversation design

Before launch, we ran test scenarios with Zorro.lv’s team to capture the brand’s tone of voice. The AI needed to sound friendly, knowledgeable, and focused on pet well-being. We reviewed and adjusted conversation examples to match how the team would naturally talk to customers.

The AI assistant was designed to handle conversations end-to-end, adapting based on customer responses and needs:

• Sends the first personalized message using order history and tone aligned with the brand
• Adjusts replies dynamically throughout the conversation
• Detects interest in product suggestions and retrieves relevant items using vectorized product data
• Asks clarifying questions when needed to improve recommendations
• Shares a personalized discount code when purchase intent is expressed
• Handles opt-out requests and removes customers from future outreach
• Flags special or complex cases for manual follow-up by the customer support team.

Technical setup

Customer data is pulled from Zorro.lv’s Shopify store, and all outreach is handled through WhatsApp. A dedicated dashboard allows the team to monitor conversations, review pre-generated messages before sending if needed, and track key performance metrics throughout the campaign to see how it’s performing.

Results

The first outreach campaign focused on lapsed customers who had previously shown strong buying intent. Results within the first few weeks:

• 31% response rate of contacted customers
• 4% conversion rate from the outreach

This exceeds typical re-engagement benchmarks. In similar campaigns, especially via email, conversion rates often stay below 1%. With personalized conversations on WhatsApp, Zorro.lv reconnected with customers in a way that felt familiar and timely. The assistant has since been extended to active customers as well, supporting ongoing engagement and repeat orders without adding pressure to the team.

Conversational commerce works because there’s something undeniably compelling about a message that feels like someone noticed you were missing. The only thing left to do? Simply being ready when the customer is.

With the right data and message, even churned customers are open to restarting the conversation. Looking to build a similar experience for your store? Let’s converse about what conversational commerce could do for your business.

The post Case Study: Conversational Commerce Brings 31% of Churned Customers Back appeared first on scandiweb.

Creating high-quality LinkedIn content regularly takes time—especially when aiming for posts that are research-backed, visually aligned with your brand, and consistent in tone. We wanted to eliminate the manual work and make the entire process operate from a simple Slack message.

How it works

This week, we built an AI Agent that automates LinkedIn post creation from a Slack channel using n8n to orchestrate the workflow. The agent combines:

• Research generation via Sonar (Perplexity API) for accurate topic context
• Post copy crafted by GPT-4o in brand voice and LinkedIn tone
• Visual assets created using GPT-Image-1 for on-brand representation
• Slack integration to deliver and iterate via chat

All of this happens from a single prompt inside Slack. The AI agent handles everything—from research to visuals—then sends back a ready-to-publish LinkedIn post, directly in the thread.

Processing and development

• n8n orchestrates API calls between Slack, Sonar, GPT-4o, and GPT-Image-1
• Sonar (Perplexity API) runs deep web research to extract current insights
• GPT-4o agents format the insights into a concise, engaging LinkedIn post and coordinate the Slack chat
• GPT-Image-1 generates visuals tailored to the post topic and tone
• Slack API provides chat interface for initiating and reviewing output

Example walkthrough

Impact

This AI agent turns a multi-hour content creation process into a 60-second Slack interaction:

• Eliminates manual research and design
• Delivers LinkedIn-ready content from Slack
• Maintains consistency in voice and branding
• Frees up creative teams for higher-level strategy

For teams seeking to scale their thought leadership without burning resources, this setup is a game-changer.

Built with

• n8n – AI automation and API workflow builder
• Sonar (Perplexity API) – web research
• GPT-4o – post copy generation
• GPT-Image-1 – image generation
• Slack API – prompt input and result delivery

Complexity

Medium

Looking to leverage modern AI tools within your company? Get in touch and explore next steps.

The post AI Agent Bites #9: AnswerRank LinkedIn post and image agent appeared first on scandiweb.

You don’t need a website visit to make a sale anymore.

More and more, shopping starts with a message: a quick question about sizing, a request for a gift recommendation, a reply to a product suggestion sent last week, or a follow-up from a customer wondering if something’s back in stock. And when the conversation feels natural and helpful, it often ends with a purchase.

That’s conversational commerce. 

It brings the buying experience into chat on any platform through assistants that engage in real time, respond with context, and guide customers toward what they’re likely to buy. Conversational commerce means dynamic conversations shaped by customer behaviour, preferences, and past interactions and overall context. 

Brands adopting this approach will reduce friction and convert interest into revenue without forcing users through long browsing sessions. Leading brands worldwide are already testing it – you can launch it! Let’s look closer at how conversational commerce works in practice, where it’s making the biggest impact, and what to consider if you’re considering launching your own.

Key takeaways

• Conversational commerce brings the shopping experience into real-time chat, helping customers find and buy products through natural 1:1 conversation.
• Success depends on clean product data, clear use cases, strong brand voice, real-time integrations, and continuous optimization.
• With the proper setup, brands can shorten buying journeys, increase AOV, and reduce support pressure without sacrificing user experience.
• scandiweb built an AI assistant for a leading Nordic DIY retailer to guide project-based shopping, leading to 95% higher AOV and a 68% increase in conversions.

What is conversational commerce, really?

Conversational commerce is a new way of selling through real-time dialogue. Instead of browsing a site, clicking, scrolling, and navigating between tabs, customers get help through a conversation. No less than one that feels personal and focused on what they actually need.

It can happen anywhere your customers already are: on your website, in a native app, over chat platforms, or even via SMS. What matters more than the channel is the experience. Your customer asks a question, gets a response, and moves forward with clarity.

Unlike traditional chatbots, modern conversational commerce tools don’t rely on scripts. Instead, they draw on customer data about past interactions with your brand and product details to guide the conversation. This makes the assistant more useful and part of the shopping journey. Done right, it’s fast, helpful, and completely integrated with how shoppers already prefer to communicate.

Why brands are investing in conversational commerce

91% of consumers globally say they’re more likely to shop with brands that offer real-time assistance. And according to industry experts, messaging will be the fastest-growing commerce channel by 2026.

Customers have grown used to getting fast, tailored responses across every platform they use. That expectation doesn’t stop at eCommerce. When someone is ready to buy or has a question holding them back, they want clear answers immediately without switching channels.

Conversational commerce shortens the time between interest and purchase and reduces drop-offs, especially for returning users. It also creates more opportunities to re-engage customers based on real data. Leading brands recognize that this means moving closer to the customer decision point without adding complexity to internal teams.

What makes it work (and what doesn’t)

AI-powered chat can increase conversion rates by up to 30% when used for guided selling. For conversational commerce to deliver such results, here’s what needs to be in place:

• The assistant has access to clean, connected product data and up-to-date info on inventory, pricing, and attributes; if it can’t answer basic questions or suggest available alternatives, the conversation breaks.
• Clear use cases and goals, like recommending products, handling repeat questions, or guiding through gift selection; otherwise, a broad, unfocused assistant often leads to more confusion.
• Assistants that respond based on real-time behavior, recent views, and purchases feel more helpful – context is everything.
• The assistant should speak in a voice that fits your brand and sound natural and human. 
• The experience has to feel instant, like chatting with someone who knows the answer before you finish asking and can suggest things you didn’t even think to ask.

Where things fall apart are generic bots with no access to product or user data and overcomplicated flows that trap the user between decisions.

Conversational commerce in practice

At scandiweb, we’ve built Concierge AI: a conversational digital assistant that feels like a boutique-style concierge. These are examples of how we see it working across different industries, based on real customer needs and typical customer profiles and shopping behaviors.

Curated product discovery

A digital beauty assistant curates full routines beyond answering your customer’s questions. After identifying the customer’s current preferences and routines, it can recommend contour, blush, lip, brush, and highlighter products that match tone and texture. Product education woven into every suggestion guides the customer through a feel-good, ultra-personalized routine, like chatting with a knowledgeable in-store rep, only faster.

Retention & gifting

When a returning customer asks for gift recommendations in his favorite watch store, the assistant instantly can reference previous orders and suggest alternatives, highlighting subtle style differences, showing product images, and offering advice on what to wear for a big event.

Multi-customer logic

A sports store customer chats in to refresh her son’s training gear. The assistant recalls previous purchases, adapts to sizing needs, and recommends tennis-appropriate shoes, even when faced with stock limitations. If a favorite item is out of stock, it can offer brand alternatives instead, tell about pricing options, and style-matching picks for mother and child, creating a helpful shopping experience (in any language, may we add).

Case study: scaling DIY sales with AI-powered guidance

Our client, a leading Nordic home improvement and DIY retailer with over 200 stores, over $1B in annual revenue, and a fast-growing eCommerce channel, serves customers across Sweden, Norway, Finland, and Denmark. 

They want to help customers who come in with a project idea, like building a deck or updating a patio, but don’t know what products, sizes, or quantities they actually need. 

Here’s the problem. 

Around 70% of annual revenue is concentrated between May and September. To support this seasonal spike, our client expands its workforce by up to 10x, primarily hiring temporary staff who need to quickly learn key product knowledge, how to talk to customers about their projects, and how to answer common in-store questions.

The training process was too slow and inconsistent to keep up. So we helped them introduce an AI-powered assistant trained on common DIY projects and linked to their full product catalog.

Project goal

Many customers walk in with a project idea but no clear understanding of what materials are required. They know what they want to build, but not the SKUs, sizes, or product combinations they need. To serve these customers better and increase basket size, this retailer needed a scalable solution to guide customers through their projects and toward purchase.

Solution

In collaboration with Algoritma, we built an AI-based ShopBot, accessible on the website. It’s trained on common DIY projects and connected to the retailer’s full product catalog. The assistant can:

• Advise on how to complete specific projects, step by step
• Recommend exact SKUs, sizes, and quantities for the task
• Provide consistent, 24/7 guidance without human staff.

How to get started

You don’t need to launch on every platform or build a fully autonomous AI from day one. But if you want conversational commerce to drive results, it needs a solid foundation.

Here’s what the process looks like:

1. Prepare your data foundation for AI

Clean product data, enriched attributes, and organized customer behavior logs give your assistant something to work with. The better the input, the more useful the conversation.

2. Build your custom conversational agent and create the interface

The interface should fit naturally into your existing flow – on-site, in-app, or embedded in chat, and be able to handle your real use cases.

3. Train your assistant to sound human and on-brand

Scripted or AI-generated, your assistant needs to reflect your tone in how it responds, how much detail it gives, and so on.

4. Connect the agent with your eCommerce & other platforms

To stay helpful, the assistant needs access to current inventory, pricing, customer profiles, and order history. Integrations with your eCommerce platform, CDP, and analytics tools make this possible.

5. Track interactions to fine-tune offers and timing

Monitor how users engage with the assistant. Where do they drop off? What prompts convert? This feedback loop lets you adjust conversation flows, tweak recommendations, and improve overall performance over time.

Start small. Test a focused use case. Then scale once the assistant proves it can make the experience faster and more helpful.

Before you know it, the best salesperson will be in your customer’s pocket.

If you’re curious how conversational commerce could work for your brand, let’s schedule a free consultation! We’ll build a tailored demo with your products and typical customer profiles to show what’s possible.

The post Conversational Commerce Explained: Be Among the First Adopters appeared first on scandiweb.

The leaders of tomorrow are already building with AI-first today. The Baltics have now caught up!

This May, at the iconic House of the Blackheads in Riga—a landmark long associated with trade and commerce—scandiweb hosted the first-ever Baltics AI Meetup dedicated to eCommerce. The energy was high, and the conversations pointed clearly in one direction: AI is becoming the backbone of how forward-thinking companies grow. 

The #1 Baltics AI Meetup sold out in advance and brought together a mix of business leaders, marketers, technologists, and AI practitioners who share a common interest in understanding how to move beyond ChatGPT tinkering and start putting AI to work in ways that actually change how businesses operate.

Attendees listened to insights and real examples and participated in open discussions about the potential and the challenges of using AI in business. We looked closely at how companies are already applying AI in eCommerce to streamline workflows and reinvent customer interactions, and what steps are needed to move from theory to practical, lasting impact.

Key takeaways

• AI is becoming the core of how modern companies operate; it’s part of the entire operations, from strategy to execution.
• Answer engine optimization is the next frontier of SEO, helping brands show up inside AI-generated answers.
• Many businesses, including startups and global brands, are already using AI to automate tasks, personalize experiences, and launch campaigns faster. 

About Baltics AI

Baltics AI (B/AI) is a meetup series dedicated to helping businesses in the Baltics explore AI in practical ways. Instead of big promises or buzzwords, each session focuses on tools, frameworks, and use cases already working in local companies today. 

Who is B/AI for?

• Business leaders & owners looking to learn how others in the region are applying AI
• Innovation leads & product managers exploring how to pilot AI in workflows and scale successful use cases across their teams
• AI-curious professionals eager to cut through the noise, hear real stories, and start experimenting with AI in low-risk, meaningful ways.

During these events, you can expect keynote speakers, real-world showcases, panels, live networking with founders, execs, and AI explorers, and actionable inspiration.

Beyond ChatGPT: real AI transformation

Kicking off the first event, Glebs Vrevsky, co-founder of scandiweb and our AI evangelist, set the tone with a bold but necessary distinction that using AI as a tool is not the same as adopting AI as a way of working.

Most teams today use ChatGPT, Gemini, or other tools to speed up isolated tasks. It’s a productivity boost, like switching from fax to email, but it doesn’t change the underlying process. That’s only the starting point.

Actual AI adoption means putting AI at the heart of how decisions are made, how teams collaborate, and how systems are built. It’s less about speeding up the old way and more about questioning whether the old way is even worth keeping.

This shift is as much cultural as it is operational. It requires new habits, new roles, and a different mindset altogether—one where AI is a core part of how work gets done.

What’s holding companies back from adopting AI?

Next, Antons Sapriko, founder of scandiweb, addressed the question, “If the value of AI is so clear, why aren’t more companies fully embracing it?”

Rather than hype or assumptions, Antons shared grounded insights based on real conversations with clients and partners. The blockers are usually strategic and organizational, not technical.

Some of the main challenges

• Fear of the unknown—many teams hesitate to adopt AI due to unclear outcomes or perceived risks
• Lack of clarity—without a clear use case or business goal, AI efforts stay stuck in the experimental phase
• Unstructured or siloed data makes it difficult to train or integrate AI tools effectively
• Overreliance on tools—teams often use AI as a productivity hack, rather than reshaping workflows around it.

For AI adoption to move beyond isolated use cases, companies need to shift how they think about change. This will most likely mean rethinking the tool stack entirely.

Getting discovered by AI via AEO

As AI tools like ChatGPT, Claude, Perplexity, and Google’s AI Overviews become key sources of answers online, traditional SEO is no longer enough. That’s where answer engine optimization (AEO) comes in.

Artyom Jurkevich, co-founder of BetterAnswer.ai, introduced the concept of AEO as a strategic approach to making content AI-readable, AI-quotable, and AI-visible. Traditional SEO and aiming for the highest rankings are being replaced by becoming the answer.

AEO focuses on helping your brand show up inside AI-generated responses by aligning your content with how these systems gather and deliver information. Artyom laid out a practical 4-step framework for AEO:

1. Know your position

Use tools like AnswerRank to find out if and how your brand appears in AI-generated answers. Identify which queries, topics, and content formats are favored.

2. Make your site AI-ready

Structure your content clearly, with headers, FAQs, internal links, concise snippets (40–60 words), and schema markup. Create pages like “Top questions about [your product]” or detailed comparisons.

3. Enhance your presence beyond your site

AI tools rely heavily on third-party signals. Building up brand mentions, positive reviews, and being referenced in trusted sources all contribute to better AI visibility.

4. Maintain your position

AEO requires ongoing refinement, monitoring where you appear, and adapting when Google or OpenAI updates their models.

AI-generated results reflect how your brand lives on the broader web, and AEO is your way of creating that perception.

AI use cases from our clients

Glebs presented a hands-on showcase of how scandiweb clients are already using AI. The use cases can be structured in three levels of adoption, each representing a deeper integration of AI into everyday work.

Level 1: out-of-the-box tools

These are plug-and-play AI solutions that improve productivity without requiring major system changes, for example:

• A custom GPT-based advisory board, simulating decision-making styles of leaders like Elon Musk and Warren Buffett
• Generating marketing assets, like product visuals and campaign ideas, using reference images and prompt-based tools
• Automating recurring content tasks, such as structuring lessons or content calendars for marketing leads.

Level 2: context-aware tools

These tools go a step further by using company-specific context to deliver more relevant output:

• Feeding transcripts from meetings and sales calls into AI to turn them into actionable insights
• Building concierge-style AI assistants that recommend products based on customer profiles and preferences
• Internal AI planners for project scoping, helping teams go from idea to structured action plan in minutes.

Level 3: connecting & automating with agents

Here, AI stops being a helper and starts acting as infrastructure. Think multi-agent systems that:

• Monitor SEO trends and auto-generate blog posts
• Act as internal procurement assistants, reading and analyzing offers
• Automate internal workflows across tools like Notion, n8n, and Claude, turning fragmented processes into smooth, AI-powered operations.

The common thread is that none of these projects started as AI initiatives. They started with a business need, and AI happened to be the best way to solve it.

Q&A: How Samsung reinvented live shopping with AI avatars

One of the standout moments of the event was the fireside chat between Boriss Sadrins (Samsung) and Maris Skujins (scandiweb), where they discussed how Samsung tackled a growing challenge in their strategy: live shopping at scale.

Traditional live shopping campaigns required large production teams, hired talent, multiple weeks of prep, and significant costs. They were engaging but not scalable across markets or frequent product launches.

scandiweb introduced a solution using AI avatars—digital versions of Samsung experts who could speak multiple languages, deliver product presentations, and scale instantly across campaigns. The avatars were paired with a custom-built live shopping engine capable of generating scripts and videos quickly without any physical production.

As a result:

• Campaigns that once took weeks could now launch in days or even hours
• Significant cost savings without compromising quality
• Flexibility to adapt to different products and audiences with no added production work
• The approach is not limited to B2C—Samsung has already applied the same system to B2B campaigns.

AI in Estonia: practical use cases across industries

To round out the day, Nikolay Sekachev, founder of NeuroSchool, gave the audience a look into how Estonian companies integrate AI into their daily operations.

Instead of big, high-risk overhauls, many Estonian teams are embracing small, targeted implementations that deliver fast results. Some standout examples include:

• AI-powered research agents using tools like Gemini to run deep competitive analysis, scan legal documents, and generate content in multiple formats: infographics, web pages, even quizzes
• Automating repetitive tasks like invoice reading, form filling, and report generation cuts hours of manual work down to minutes
• Data enrichment for eCommerce, linking product records with real-time supplier data to create better listings and reduce mismatches
• Vision-based diagnostics, including tools that analyze facial or skin images to provide feedback in medical, wellness, or education contexts
• Voice-powered assistants built to help with schoolwork or even car buying

What makes these examples compelling is their variety and how attainable they feel. No giant AI departments or multi-year transformation projects; just sharp use of the available tools, guided by real business needs.

Closing thoughts

AI has already changed how businesses solve problems, create content, interact with customers, and make decisions. Global brands like Samsung and agile teams in the Baltics and beyond are moving forward—sometimes with big leaps, often with small, smart steps. The shift is underway, whether through plug-and-play tools, contextual assistants, or fully automated agents.

There’s no one-size-fits-all path, but all businesses are learning that real value comes not from using AI to speed up old workflows but from rethinking them entirely.

At scandiweb, we’re continuing to explore what this means for eCommerce and are excited to help others do the same.

If you’re ready to build something AI-first or just want to discuss the possibilities, we’re here. Send us a message, and let’s come up with something great together!

The post The AI Era in the Baltics Is Officially Underway: Baltics AI eCommerce Meetup #1 Highlights appeared first on scandiweb.

Creating quick prototypes or landing pages typically involves multiple steps—design, coding, and in some cases deployment—which can slow down idea validation and experimentation.

How it works

This week, we experimented with Google Stitch to generate a mock landing page from simple natural language prompts. We asked Stitch to create a landing page for a watch shop called WorldTime.net. Stitch responds with mockups and allowed us to refine the design interactively through chat.

The standout feature? Stitch provides copyable code after every response, enabling seamless handoff.

Processing and development

We copied Stitch’s generated code directly into Cursor, which created a Next.js project and implemented the design. After some tweaks and experimentation, we ended up with a fully animated, polished landing page.

Finally, the project was deployed effortlessly on Vercel, delivering a live site—all powered by natural language prompts.

Visit the landing page yourself here.

Output

• A fully coded, animated landing page for WorldTime.net
• Deployment to Vercel, ready for live testing and user interaction

Example walkthrough

Original mockup by Stitch

Code provided by Stitch

Final landing page via Cursor

Impact

This workflow drastically reduces time from concept to live product—letting teams prototype and deploy web pages faster than ever without manual coding.

Built with

• Google Stitch – AI-Powered website and app mockups
• Cursor – AI coding environment that automates project setup and implementation
• Vercel – Easy deployment platform for Next.js projects

Complexity

Low

Looking to leverage modern AI tools within your company? Get in touch and explore next steps.

The post AI Agent Bites #7: Building a Landing Page with Google Stitch and Cursor appeared first on scandiweb.

Creating high-quality, SEO-optimized blog content is often a time-consuming and fragmented process. Gathering trending keywords, generating engaging copy, creating relevant images, and organizing drafts typically require multiple tools and manual coordination. This complexity slows down content teams and limits their agility in responding to market trends.

Our solution streamlines the entire blog creation process by integrating keyword research, content generation, and draft management into a single, seamless AI-driven workflow.

How it works

MCP integration with Claude transforms Claude into a powerful blog-generating assistant capable of gathering all necessary context to produce high-quality blog content. The MCP connection enables Claude to automatically fetch trending keywords from AHREFS, generate blog text and images based on those keywords, and submit completed drafts directly to Notion—all through a natural, conversational chat interface.

Processing and development

• MCP: This seamless workflow is powered by MCP (Model Context Protocol), an open standard developed by Anthropic to enable AI models to securely access and interact with external tools and data sources (API’s).
• N8N MCP server: The MCP connection is powered by an N8N workflow that orchestrates all API calls. N8N enables automated fetching of trending keywords from AHREFS, triggers image generation requests, and writes the generated content into Notion, acting as the backbone of the entire process.
• Claude Desktop: Anthropic’s Claude lets us use natural language prompts to guide blog creation. Its MCP integration allows us to feed highly specific, curated context (keywords, briefs, and style guidelines) into Claude. Claude then produces interactive artifacts including keyword suggestions, detailed content plans, images and fully formed blog drafts ready for review.

Output

• A fully automated blog generation workflow combining keyword research, AI copywriting, and image creation.
• Blog drafts organized and stored in Notion, ready for editing and publishing.
• Visual and textual content creation all accessible through an intuitive chat interface.

Example walkthrough

Impact

What you just saw was accomplished using a single chat with Claude—from keyword research to content planning to a full blog draft. MCP allows for integrating any API into Claude, allowing chatbots to access the most relevant context possible. The possibilities are endless and allows for simple chatbots to become fully-fledged Swiss army knives of productivity.

Built with

• N8N – AI workflow automation platform managing API orchestration and content flow
• Claude Desktop – Anthropic’s Claude desktop app with MCP integration for natural language-driven content generation and context management
• AHREFS API – For trending keyword data powering targeted blog topics
• GPT-Image-1 via OpenAI API – For generating blog images

Complexity

Low

Looking to leverage modern AI tools within your company? Get in touch and explore next steps.

The post AI Agent Bites #6: Claude Blog Factory using MCP appeared first on scandiweb.

We saw this customer service agent demo in the Google Cloud Next 2025 Keynote. Looked interesting and we wanted to try it out!

Google has provided the “adk-sample” repo to showcase their latest Agent Development Kit, containing various AI agent “samples” built using the ADK. We went ahead and tried out the Customer Service sample, to see how it compares to the keynote demo.

You are given a role of the customer in this sample agent, so we decided to test how it can help revive a neglected orchid from my apartment corridor.

This agent was tested using the ADK web UI, which is part of the Agent Development Kit. The UI provides the means to locally; chat with the agent, manage chat sessions and see past “events” which include a visual breakdown of the tool calling made by the agent. The UI also includes artifacts (for uploaded files) and the “eval” tab for testing various scenarios, but neither of those seemed to work in our case. The screenshot above also shows the customer profile object, which I edited to personalize the experience!

Walkthrough

After instructing it to add the suggested products to the cart, we asked about the price and discount. The agent wasn’t able to give us the price of the cart, but gave us a discount!

Conclusion

As you can infer, the agent wasn’t quite what we saw in the keynote demo—something Patrick Marlow himself acknowledged.

Notably, some functionality like the Live API was missing. When we tried using live video or audio (as shown in the keynote), an error appeared indicating the default model didn’t support the Live API. Switching to a model that does support it led to a different error instead.

Finally, the Agent Development Kit offers a solid development environment with a built-in Web UI for testing. However, more work is needed to bring keynote functionality to production. The Web UI is strictly local, and deploying the agent to Google Cloud’s Vertex Agent Engine means local tools won’t carry over—they must be manually configured. Likewise, the Web UI is only for local testing; a custom interface is needed for production.

Looking to leverage modern AI tools within your company? Get in touch and explore next steps.

The post AI Agent Bites #5: Agent Development Kit – Customer Service Agent walkthrough appeared first on scandiweb.