Estimated reading time: 18 minutes
Your competitors are not waiting for the perfect AI strategy. They are starting. Small or big. They are deploying, learning, and iterating. They are quietly building advantages that compound over time.
AI projects average 3.7x return on investment (IDC Study), yet most enterprises are still debating how to start or how to scale. The organisations in this article decided to act. What follows are more than 15 real AI and agentic AI use cases built by Devoteam‘s teams across EMEA. From hospitality and luxury fashion to financial services, media, and manufacturing. Every result is real. Every number is sourced. Every use case is one to consider for your business.
More than 15 AI & Agentic Use Cases Built by Devoteam
AI projects average 3.7x ROI. While competitors wait, European leaders are already deploying. Explore our replicable cases across EMEA.
Choose Your Sector
Discover proven AI transformations within your industry.
Results
Hospitality
1. Strawberry Hotels Saved 36,000 Hours a Year with a Single Agentic AI Deployment
The problem: Strawberry (formerly Nordic Choice Hotels) is one of Europe’s largest hotel groups with 230+ properties and around 20,000 employees. They had a sprawling documentation problem. Decades of system upgrades across hundreds of hotels had created an ocean of procedures, policies, and knowledge. The real issue? Staff could not efficiently find what they needed. Every unanswered question became a support ticket or a wrong decision.
What Devoteam built: Working with Google Cloud, Devoteam engineered ‘Scout AI’. Scout AI is an enterprise-grade agentic AI assistant using a secure Retrieval-Augmented Generation (RAG) architecture. It integrates directly into Google Chat without requiring data migration or another platform. Devoteam deployed the agentic AI framework using Vertex AI Agent Builder. Once the main agent was running, the team tuned individual “Sister Agents”, such as the guest-facing agent Lykke, to match the distinct persona of specific hotels. A continuous telemetry loop was also established. If multiple users searched for a topic that returned no answer, the system immediately alerted the marketing team to update the documentation.
The results: 25,000 hours returned to employees, who now resolve queries in seconds instead of minutes. 11,000 hours returned to guests through expanded AI-powered 24/7 guest support. 98% positive guest sentiment, driven by the accuracy of the RAG architecture. Staff adoption skyrocketed when Scout was given a deliberately friendly, “sassy” personality. Proof that change management and UX design matter as much as the underlying model.
Read the full story: how Strawberry saved 36,000 hours a year with Scout AI.
Retail, Luxury & Fashion
2. Hypebeast Cut $3 Million in Annual Costs with AI-Powered Operations
The problem: Global streetwear and culture platform Hypebeast had critical information locked in siloed systems. Their inventory data sat in Excel spreadsheets, editorial workflows were managed manually, and customer intelligence was scattered across tools. The result: slow decisions, operational drag, and a growing gap between the business’s ambitions and its operational reality.
What Devoteam built: Devoteam implemented Gemini Enterprise (formerly Google Agentspace) with a phased adoption strategy. We first introduced teams to simpler, high-value tasks like document search, multimodal content summarisation, and email generation. The team then focused on more complex tasks. Merchandising was transformed by uploading Excel spreadsheets containing real-time inventory insights into a centralised Google Drive connected to Gemini Enterprise. This enabled staff to ask natural-language questions, such as “Where is this pallet of sneakers today?” and receive instant answers. An automated editorial workflow was also built for the content team.
The results: $3 million in annual cost savings. The implementation required no platform overhaul. Hypebeast’s staff continued using Excel and familiar tools while the AI layer handled complex information retrieval behind the scenes.
Read the full story: how Hypebeast cut $3M with Gemini Enterprise.
3. DREST Reduced Fashion Content Creation Time by 99% Using AI
The problem: DREST is a luxury fashion styling app featuring 300+ brand partners. They employed five full-time designers to manually edit clothing assets onto model images using Photoshop-like tools. Every image took approximately half a day per person. Training a new designer took six months.
What Devoteam built: Through the AWS-funded ADAPT programme, Devoteam validated AI feasibility and built “Dresty”. Dresty is a self-service prototyping tool allowing DREST’s content team to test AI and pre-trained ML models for virtual try-on at scale. The evaluation covered multiple models to find optimal approaches for different clothing types, with specific success criteria around texture fidelity and fitting quality.
The results: The content creation bottleneck disappeared. Processing time dropped from half a day per image to under one minute. The solution was delivered on time and within budget, with the DREST team receiving a tool that can evaluate new AI models as they became available.
Read the full story: how DREST cut image creation time by 99%.
4. Longchamp Went from 10% to 98.6% Copilot Adoption in Six Months
The problem: Luxury house Longchamp started a Microsoft Copilot pilot with 50 licences, 40 assigned… And a usage rate of around 10%. No strong executive sponsorship. Limited budget. A fast-growing company with overloaded teams that had no time to experiment with new tools.
What Devoteam built: Devoteam’s approach was counterintuitive: rather than rolling out broadly, they built a restricted, exclusive community of early adopters that created internal desire. Concrete training around specific, high-value use cases, meeting minutes generation, RFP preparation, translation, presentation creation, turned sceptics into advocates. The “Leagues” approach maintained exclusivity while optimising paid licence allocation against budget constraints.
The results: 98.6% user satisfaction. 3.4 hours saved per week per user. Exploding internal demand for expanded licences. The project is now scaling into Copilot Studio for custom agent development.
Read the full story: how Longchamp went from 10% to 98.6% Copilot adoption.
Financial Services & Insurance
5. Portugese Bank Built a Knowledge Management Chatbot That Deploys in a Single Day
The problem: A Portuguese bank needed to enable its employees to instantly access internal knowledge across multiple organisational domains — HR, legal, compliance, product — without depending on specialised teams to answer every question.
What Devoteam built: Devoteam created a flexible wizard application that enabled this bank to generate internal chatbots within a single day, capable of interacting with knowledge bases across different departments. The solution enhanced organisational efficiency and knowledge sharing, letting employees access topic-specific internal knowledge within seconds and significantly reducing time spent searching for information.
The results: Employees became capable of independent problem-solving across domains. Onboarding time for new staff shortened. The knowledge management burden on specialist teams dropped substantially.
Watch the story: Reshaping Banking with AI
6. MBA Group Accelerated Client Onboarding with Generative AI
The problem: MBA Group, a financial services firm, faced a time-consuming, document-intensive client onboarding process — a classic bottleneck where compliance requirements, manual data extraction, and document generation were creating delays and consuming advisor capacity.
What Devoteam built: A generative AI solution on AWS that automated document data extraction and client-specific document generation, reducing the time and human effort required per onboarding while maintaining full compliance with regulatory requirements.
The results: Client onboarding time reduced significantly, with advisor capacity freed for relationship-building rather than document management. The solution demonstrated that heavily regulated, document-intensive processes are not barriers to AI — they are prime candidates for it.
Read the full story: how MBA Group accelerated onboarding with GenAI.
7. BPCE Assurance Trained 900 Employees on Generative AI — and Discovered Use Cases Their Competitors Are Still Trying to Find
The problem: When the BPCE Group integrated “MaIA”, a privatised version of ChatGPT, into its work environment, the Personal Insurance business of BPCE Assurances faced two urgent questions: how to acculturate 900+ employees to the use of generative AI, and how to build real experience fast enough to stay at the innovation frontier. Without structured guidance, AI tools risk becoming shelfware, adopted in name, unused in practice.
What Devoteam built: Devoteam supported BPCE Assurances in designing and running a comprehensive generative AI acculturation and training programme for its entire Personal Insurance workforce. The programme covered prompting methodologies and best practices, and was structured to identify both cross-functional and business-specific AI use cases from within the teams themselves, rather than imposing them top-down. As Nofel Goulli, Deputy Managing Director at BPCE Assurances, described it: the ability of Devoteam’s team to evolve and adapt quickly — considering feedback and adjusting the rollout accordingly — was what turned ambition into concrete achievement.
The results: 900 employees trained across the Personal Insurance business. Cross-functional and domain-specific GenAI use cases identified and validated from within the workforce — giving BPCE Assurances a living pipeline of AI opportunities grounded in operational reality rather than executive aspiration.
Read the full story: how BPCE Assurance trained 900 people on GenAI.
Telecommunications
8. Bouygues Telecom Is Using GenAI to Recommend What Your Family Watches Tonight — and Built an Internal ChatGPT Your Competitors Haven’t Heard Of Yet
The problem: Bouygues Telecom’s innovation lab had a mandate to build the next generation of customer and employee experiences using generative AI — but needed a partner with both deep cloud expertise and frontier-level AI knowledge to go from idea to production-ready product at the speed the market demanded.
What Devoteam built: Two parallel GenAI deployments. The first was “Ensemble ce soir”, a family TV content recommendation service powered by generative AI, built in collaboration with AWS, that finds the ideal content for any family to watch together in under five minutes. According to Laurent Sauvage, head of Bouygues Telecom’s Innovation Lab: “Working with a partner with advanced mastery of the Cloud environment, particularly AWS, and strong expertise in AI and RAG topics is precious. This combination of Cloud and AI expertise makes for very successful support.” The second was Lucia, a privatised internal ChatGPT built on OpenAI APIs with dedicated cloud hosting, guaranteeing the confidentiality of all employee interactions. A real-time in-store translation service was also deployed, allowing advisors and customers to converse in their own languages simultaneously via smartphone, launched in a flash-code format requiring zero installation.
The results: Three production-grade GenAI applications deployed — a customer-facing product, an internal productivity tool used daily across the company, and a real-time translation service for the store network. As Sauvage noted, Devoteam’s support helped Bouygues Telecom “build scalable and robust architectures, go into production, and therefore accelerate the time to market.”
Read the full story: how Bouygues Telecom shipped GenAI products with Devoteam.
9. Cellnex Went From Copilot Pilot to Measurable Productivity Gains — with Quality of Work Up and Effort Down
The problem: In 2024, Cellnex,Europe’s leading operator of wireless communications infrastructure, needed to facilitate a smooth and effective transition to Microsoft 365 Copilot, support users throughout the process to ensure the success of the change, and maximise user satisfaction with the tool. Like most large enterprises, the challenge was not access to the technology. It was making people actually want to use it — and use it well.
What Devoteam built: Devoteam designed a full Copilot Adoption and Change Management Plan for Cellnex, including a tailor-made communication strategy. Before the pilot launched, Devoteam ensured all technical prerequisites were met. Training materials and resources were created and delivered to users throughout the pilot, with continuous support and guidance provided to every participant. Usage and performance data were monitored and measured throughout to verify whether objectives were being achieved.
This engagement was subsequently cited as the reference project when Devoteam achieved the Microsoft Copilot Specialization in October 2025, one of the first partners globally to hold this recognition, demonstrating verified expertise across Microsoft 365 Copilot Chat, Copilot Studio, and Agents.
The results: Cellnex users concluded the pilot recognising the significant potential of Copilot. Measurably increased productivity. Improved quality of work, achieving better results with less effort, leading to faster and more effective ways of working. High satisfaction with the tool, demonstrating that the adoption process was carried out effectively. Devoteam’s engagement with Cellnex continues, with ongoing support to maximise the value of the full Microsoft 365 platform.
Read the full story: how Cellnex built a Copilot adoption plan that stuck.
Public Services & Logistics
10. Portugal’s Post Office Raised Customer Satisfaction by 40 NPS Points with a GenAI Chatbot
The problem: CTT, Portugal’s postal service, faced an overwhelmed call centre. 50% of all interactions concerned a single question: Where is my parcel? Their existing menu-based chatbot provided only generic responses, leaving customers frustrated and agents buried in routine queries.
What Devoteam built: Devoteam developed “Helena”. Helena is an advanced transactional chatbot powered by Microsoft Azure OpenAI Services, designed to engage in natural language conversations and integrate with multiple data sources via APIs for accurate, real-time parcel tracking. Around 70% of the project effort was dedicated to ensuring Helena’s tone of voice matched CTT’s brand precisely. A lesson in why AI adoption is as much an editorial challenge as a technical one.
The results: A 40-point increase in Net Promoter Score, 60% more daily interactions handled, and over 281,000 responses delivered in just the first three months. The solution reduced call centre volume while improving customer satisfaction.
Read the full story: how CTT Portugal built Helena with Azure OpenAI.
While Helena demonstrates the power of transactional chatbots, the strategy behind these deployments requires a dedicated approach. For a deeper look at how to implement these technologies within support teams, explore our comprehensive AI agents customer service guide.
Media & Publishing
11. L’Équipe Produced AI-Generated Olympic Podcasts
The problem: French sports media outlet L’Équipe faced a tight 30-day deadline to deliver podcast episodes covering the Paris Olympics, with its entire editorial staff already committed to daily coverage of the Games.
What Devoteam built: Using Amazon Bedrock, Devoteam created a solution that transformed L’Équipe’s existing web content into audio podcasts in a repeatable, scalable process. The AI did not generate original content — it used the journalists’ published work, fed through Bedrock, with AWS Lambda automating text validation and text-to-audio conversion, and AWS Step Functions orchestrating the full workflow. The tool was designed to be simple enough for the L’Équipe team to operate independently after an initial briefing.
The results: All 19 episodes delivered on time, at broadcast quality, without diverting a single journalist from their Olympic coverage duties. The process is now repeatable for future major events.
Read the full story: how L’Équipe shipped AI podcasts
Manufacturing
12. DYMO Is Building the ‘Factory of the Future’ on a Modern Data Platform
The problem: DYMO, the global labelling products manufacturer, needed to move from a fragmented, legacy data environment to a modern, AI-ready data platform.
What Devoteam built: A cloud-native data platform designed specifically to power AI use cases across DYMO’s manufacturing operations, from predictive maintenance to operational analytics. The platform provides the governed, high-quality data foundation that production-grade AI requires, built to scale as DYMO’s AI ambitions expand.
The results: DYMO now has the infrastructure to deploy AI use cases across manufacturing operations — moving from “data as a by-product” to “data as a competitive asset.”
Read the full story: how DYMO built an AI-ready data platform.
Professional Services & Research
13. Source Global Research Automated Research Intelligence with 90%+ Accuracy
The problem: Source Global Research, a London-based research and strategy firm for the global professional services sector, saw an opportunity to enhance how it turned raw data into actionable insights. Three areas stood out: qualitative research analysis, entity recognition across company names and key terms, and the firm’s rigorous but time-intensive thought leadership evaluation methodology. The challenge was scaling analytical output without scaling headcount.
What Devoteam built: A phased engagement funded through AWS programmes. The first phase (April–May 2025) was a six-week proof of concept delivering three AI tools: a Named Entity Recognition pipeline combining AWS Bedrock with serverless pre- and post-processing, achieving above 90% extraction accuracy; a Qualitative Summarisation Engine built on AWS Bedrock’s large language models, capturing metadata tags at 70%+ accuracy alongside text summaries; and a Thought Leadership Evaluation Tool using Amazon Textract for document processing and four specialised AI agents to score publications against Source’s proprietary analytical framework. The second phase (August–September 2025) was a production readiness assessment under the AWS RAPID programme, covering usage analysis, prompt engineering, governance, and SLA benchmarking — delivering a full implementation roadmap for deployment at scale.
The results: The Named Entity Recognition pipeline significantly exceeded expectations at above 90% accuracy. The Thought Leadership Evaluation Tool enables baseline AI scoring across thousands of documents, freeing analyst time for the manual review of featured content. AWS funding reduced the capital outlay, letting Source validate technical feasibility and business value before committing to full-scale production. In the words of James Foden, Head of Offerings at Source Global Research: Devoteam’s phased approach “gave us the confidence to double down on the use case with the strongest impact and progress it further, without committing too early.”
Read the full story: how Source Global Research scaled research with AI.
14. Version Two Scaled Event Intelligence Insights with AWS Bedrock in a Five-Week PoC
The problem: Version Two is the UK-based SaaS company behind Evessio, a cloud-native platform powering awards, conferences, and virtual events for publishers and enterprise organisers. The team had started experimenting with ChatGPT to analyse event data and cut down manual processing. The workflow worked, but it was slow and it did not scale. With an MVP launch approaching, they needed a secure, AWS-native GenAI solution that would integrate with their multi-tenant architecture — not another bolt-on.
What Devoteam built: A five-week proof of concept delivered through the AWS-funded ADAPT programme, built for Version Two’s Co-founder and CTO Paul Schnell. Using Amazon Bedrock foundation models, Devoteam designed a reusable, modular summarisation framework that generates intelligent, contextually aware reports across different competition formats. The architecture was shaped around Version Two’s specific use cases but structured to scale — with a forward-thinking call structure and downstream integration paths baked in from the start.
The results: The summarisation framework is being embedded directly into Evessio’s new platform as part of the November MVP. Significant reduction in manual workload across event cycles, faster time-to-insight with automated report turnaround, and tighter platform governance via Bedrock-managed models. As Paul Schnell put it: “This project has been instrumental in kickstarting a key enhancement to our platform.”
Read the full story: how Version Two scaled event intelligence with Bedrock.
Technology, Software & Gaming
15. Empire Games Is Using LLMs to Generate Mobile Game Levels Automatically
The problem: Empire Games, a London-based gaming studio behind titles like Nuts & Bolts Screw Wood Puzzle and Empire Bingo, was creating playable puzzle levels manually. The process was slow and hard to scale. The team needed a pipeline that could algorithmically generate valid puzzles from defined input parameters — reliably, with minimal human intervention, and robust enough to be low-friction for both developers and end users.
What Devoteam built: A proof of concept in collaboration with AWS: a fully LLM-driven workflow handling every stage of gameboard generation from start to finish. Python-based orchestration integrated with Amazon Bedrock. Claude’s thinking tokens interpret the initial input and guide subsequent requests, ensuring each gameboard aligns with the intended puzzle design. Tool integrations execute specific operations with precision to reduce errors. An agentic approach preserves context across iterations. Every gameboard is validated against a defined schema before delivery — a safeguard for quality and playability that Empire Games’ founder would later cite as the most important lesson from the project.
The results: A 70% usability rate across the initial set of generated gameboards — proving both feasibility and clear progress toward an efficient solution. Automated generation was demonstrably more cost-effective and scalable than manual level design. As Alex Palaghita, Founder & CTO at Empire Games, put it: “For other companies facing similar challenges, my advice would be to focus on building a clear validation pipeline early. Automating content generation is only useful if the outputs can be trusted.”
Read the full story: how Empire Games generates levels with LLMs.
16. Cegid Deployed a 24/7 AI Support Agent Handling 500,000 Annual Tickets
The problem: Cegid, a major European software publisher, faced the challenge of efficiently managing information and usage requests across its portfolio of 90 solutions , representing around 40% of its 500,000 annual support tickets.
What Devoteam built: Working with Cegid and Google Cloud, Devoteam deployed an intelligent conversational chatbot using Dialogflow, selected for its advanced AI capabilities, mature generative AI ecosystem, and commitment to responsible AI. The implementation includes sentiment analysis to monitor customer emotions and enable proactive service improvement.
The results: 24/7 AI-powered support across Cegid’s full solution portfolio, with sentiment analysis providing ongoing service quality intelligence that no human-only support operation could generate at scale.
Read the full story: how Cegid automated 500,000 tickets a year with Dialogflow.
Devoteam’s Own Agentic AI Builds
17. RegulAIt: An Agentic AI That Turns Regulatory Change into Assigned Tasks — Not Another Summary
The problem: Regulatory velocity has outpaced human processing capacity. Compliance teams spend 60–70% of their time on manual research, reading new publications, amendments, and court rulings across multiple jurisdictions. Traditional monitoring tools alert you that something changed. Standard AI chatbots summarise the text. Neither tells you what to do about it. Your analysts still have to read the summary, decide what it means for your business, and manually create the follow-up tasks. You have scaled the conversation, not the operation. Critical changes get caught too late.
What Devoteam built: RegulAIt is an agentic AI platform that executes the compliance workflow rather than just describing it. Autonomous agents continuously crawl regulatory websites, legal databases, and official sources against a custom taxonomy. When a new document is detected, agents extract obligations, assess business impact against uploaded internal policies, and generate specific action items — assigned to the right department, with priority, timelines, and a full audit trail. The reasoning is visible in real time, not hidden in a black box. A Regulatory Delta Analysis feature compares new amendments against existing requirements. Built on open-source libraries and designed to sync with existing GRC systems, with ServiceNow as a strategic partner shaping the roadmap.
The results: Compliance shifts from a sequential, human-gated process to a parallel, continuously operating system. What used to take weeks of analyst time happens in minutes. A Friday-evening regulatory publication in Germany is assessed, mapped to hedging strategies, and turned into assigned tasks by Monday morning. A court ruling reinterpreting disclosure requirements is flagged, analysed, and routed to underwriting within hours instead of weeks. Teams move from “what changed” to “what we must do” without the manual bridge in between.
Read more: inside RegulAIt, agentic AI for regulatory compliance.
The Pattern Behind Every Success Story
More than 15 use cases. Nine industries. Four technology platforms. One consistent pattern.
None of these organisations waited for perfect conditions. None of them completed a two-year AI strategy process before writing a single line of code. They identified the use case with the strongest impact, validated it rapidly, and moved to production. How? Simply by following & trusting the methodology that Devoteam’s 1,000+ AI consultants and 800+ successful AI projects have refined across EMEA.
The organisations that are building AI advantages today are not smarter than yours. They made a decision to start. And they chose a partner with the engineering depth to take them from idea to production. Not just from idea to slide deck.
As Devoteam’s leadership has stated clearly, the biggest risk to organisations today is not deploying powerful technology without a strategy. It is not starting at all.
How would you like to start your AI transformation?
Replicate the success of Europe’s market leaders with a tailored discovery path.
Where do you want to go?
Select your preferred technology
Recommended Architecture
Success Story: Strawberry Hotels used Agentic Search to solve their documentation ocean.