Contextual AI – Experience like no other

Strategi

Every store visit, every online interaction, is designed to bring more inspiration, creativity, joy and aesthetics into your daily life. Together with an experience that feels personal and thoughtfully curated for the customers' demands, we create “experiences like no other”. But do you know why the customers visited you in the first place? In what everyday context or need they decided to visit your brand? When covering more than 150 different product categories and many more entry points of interest, the answer was difficult to find. We are known for a unique content universe and a persistent and recognisable tone of voice, and exact content is a vital way to bond and create customer relations that last over time. We wanted to combine digital touchpoints, and customer demands, to create hyperpersonalised contextual content to re-create their first point of interest and re-engage them.

But across digital touchpoints, there are challenges: What happens when customers search for products on the web shop and suddenly stops the session? A crucial touchpoint was lost. What triggered the abandonment when customers browse the website and then suddenly disappear? No one knows. But more important, how can we re-create that state of mind that urged the visit in the first place and get the customer back in the loop? This is where AI became the key to a revolutionary and innovative personalisation strategy.

We call it ‘Contextual AI’. A solution that captures customers' interest and behaviour and combines innovative technology with owned first party data. Tailored and personalised to enhance the customer experience across 15 markets and 7 languages.

The solution gathers online behaviour from both the app and website, tracking onsite searches, product and category browsing, DIY interactions, clicks, favourites marking, and basket activity. This data is used to create a contextual profile, triggering an automated flow that generates hyperpersonalised AI-driven emails with tailored product recommendations and DIY content.

Furthermore, we created a continuous feedback loop to improve the AI-solution. By combining few-shot learning and fine-tuning based on the performance of each uniquely and identified AI generated e-mail, we constantly optimise the performance in real-time.

Løsning

The key strength of this solution lies in the robust tech setup that we, together with IMPACT Commerce, has built since starting their journey together in 2019. This solid foundation has enabled the seamless implementation of AI-powered solutions making personalisation a reality across multiple markets while ensuring cost efficiency.
Now, when a customer abandons their cart, shows interest in a product or category, or abandons a search, Relewise triggers an automated email flow. Relewise then sends the relevant user data to our AI engine, which generates highly personalised content including the subject line, preheader, headline, body, and CTA tailored to our unique "Anna & Clara" storytelling. This AI-optimised content is then returned to Relewise, which forwards it to Salesforce Marketing Cloud for final assembly and delivery. The result is a seamlessly orchestrated, hyperpersonalised email experience that re-engages customers at the perfect moment.
The solution started with an AI solution for low-funnel flows, specifically product, category, and abandoned cart, using Søstrene Grene’s internal AI engine. This engine utilises state-of-the-art large language models and follows a modular design, allowing effortless upgrades without the complexity of a full-scale AI rebuild.
Then, we introduced an additional layer of complexity by expanding the solution to include abandoned search. This enhancement incorporated complex, asynchronous AI flows for seamless real-time processing. To achieve this, we integrated IMPACT’s AI accelerator, a serverless AI platform, with pre-built infrastructure components and a flexible API which enabled seamless integration into Søstrene Grene’s existing tech stack.
We have fine-tuned our tone of voice using existing data and collaborated with our internal communication team to ensure consistency with our brand identity. To drive continuous improvement, we have implemented a feedback loop that analyses customer actions to assess the relevance of our personalized emails. This data enables us to refine our communication style based on customer preferences, creating a system that becomes increasingly effective over time.

This means that whenever a registered customer interacts or searches for a product but doesn’t complete a purchase, AI evaluates the search results and determines the best follow-up action: 
• If strong matches exist, AI generates a hyperpersonalised that speaks directly to the customer’s interests.
• If no perfect match is found, a fallback message ensures the email remains engaging and relevant. 
• If the search returns no results, the system steps in to enhance the search experience and automatically corrects spelling errors, suggests synonyms etc. to generate more accurate results.