Directions EMEA 2026 in Paris made one thing clear for the Business Central community: AI is no longer a side project. Almost every session, from Microsoft's own keynotes to the MVP deep dives, came back to Copilot, agents and what happens to an ERP when people stop opening it to get their answers.
The BC4 team came home with different priorities depending on where we sat. Retail technology, operations and marketing each pulled out their own themes, but several came up again and again. Here are the ones we think matter most to retailers and sports organisations running Business Central, or thinking about it.
Between sessions we spoke to colleagues from Microsoft, LS Retail, Adyen, Revolent and more.
Microsoft announced Business Central in Copilot Cowork, with a partner preview due in November and December. The practical change is simple: a user will be able to ask a question and get an answer without opening Business Central at all.
For retailers, the questions are where it gets interesting. "Which products are selling out in some stores but sitting unsold in others?" or "How did last weekend's sales compare with last year, by store?" are the kind of questions that today need a report, a filter, or a call to the one person who knows where to look.
Several sessions pointed at the same bigger shift, often described as headless ERP. As AI agents work across systems, the interface becomes less central and the value moves to what sits underneath: the business logic, the trusted data and the connected processes. For a retail estate with stores, ecommerce, warehouses and finance all feeding one picture, that foundation matters more than any single screen.
"As agents start to work across our systems, the interface matters less and what sits underneath matters more: the business logic, the trusted data and the connected processes. For businesses, that is where the long term advantage will be built."
Bevan Miller, Operations Director, BC4
The wider transition is still developing, and nobody should rip anything out on the strength of a keynote. But the direction is clear enough to plan for. If more of your people will use AI tools rather than log in to look at data directly, it is worth deciding now how your business will talk about, and design for, that way of working.
Plugging AI into Business Central is the easy part. Getting good answers out is harder, because in most businesses the data is messy, the codes are not explained and the definitions have never been agreed.
Take the store comparison question above. If "sales" means something different in stores and online, or item categories have been set up differently over the years, the AI will still give you an answer. It will just be a confident answer sitting on shaky ground.
The lesson from Paris is that data readiness is the real starting line for AI. Before switching anything on, it is worth asking three plain questions. Would someone new understand your item and location codes without a guide? Have you agreed what your core measures mean, such as sales, margin and stock availability? And is your master data clean enough to trust? Our consulting and advisory team can help you work through them.
Business Central licences already include Copilot features. These include bank reconciliation, chat, analysing lists, marketing text for items, and turning e-documents into purchase lines. Most customers have not switched them on.
That makes them the cheapest place to start. Pick one workflow your team does every week, enable the relevant feature in a test environment, and see what people make of it. It builds confidence, shows where your data needs attention, and gives you real examples to learn from rather than a slide deck full of promises.
It is worth checking which Copilot features are available and enabled in your own environment, as availability depends on your version and set up.
Partners can now build agents inside Business Central for specific tasks, test them properly and see what each run costs. That last point matters. It turns agents from an interesting demo into something a business can budget for.
The sensible place to start is small and repetitive. Look through your service desk tickets for the tasks that come up again and again, then pick one or two that an agent could handle. In retail, that might be routine queries that follow the same steps every time. Start narrow, measure what it saves, and expand only when the results justify it.
Agents are not only a Business Central topic either. AI can help people in every department work more efficiently, and in many businesses some people already have agents running. The risk is that they grow unevenly, one experiment at a time. A clearly defined strategy for where and when to use agents will do more for you than a scatter of good ideas.
AI tooling is moving fast, and so are the commercial terms behind it. Prices change, and providers can disappear. Whatever you build, avoid tying your processes to a single AI provider. You should be able to switch to a different model quickly, with minimal disruption to day to day operations. It is a fair question to put to any partner building agents for you: how portable is the work?
On the development side, the focus has shifted from producing code to checking it. Microsoft and the MVP community showed tools to help, including BCQuality, a shared library of common AL mistakes that AI learns from, and AI code review in AL-Go. A person still reviews and signs off. The AI simply does more of the work.
"Writing code is no longer the slow part. Making sure the AI writes the right code is. The teams that get the most from AI will be the ones that write the test first and prove every change, not the ones that simply generate the most code."
Nuno Soares, Retail Technology Director, BC4
The case for testing was the most practical part. Microsoft tested AI on real AL bugs, and on bigger fixes it got the answer right only about half the time. The code compiled, but the logic was wrong. Microsoft's session treated tests as part of the development loop (write the test, build, run, fix, repeat), and Vjeko went further by suggesting teams write the test first, because it forces a decision on what the code should do before the AI writes it.
Our development team wants to take that one step further with spec driven development, where the specification and the tests are written before any code is started. The spec says what the change must do, the tests prove it, and the AI then works against both rather than guessing at intent.
The tooling is catching up. Testing is getting simpler in VS Code, pipelines and AI can now run tests on their own, and AL Runner, an MVP tool, can run thousands of tests in seconds and show whether an extension breaks standard Business Central.
If you are buying development work, this is a fair question to ask any partner: how is AI-assisted code tested before it reaches your system? Do they work to a shared development framework, with the same standards, skills and tests applied by every developer and their AI agents, or does each person do their own thing? Faster delivery only helps if the change is right.
Not everything was technical. Three people themes stood out.
The first was skills. Microsoft talked about becoming its own "customer zero", using AI internally before asking customers to. That fits a view we share: teams need practical chances to use AI, a way to share what works, and time to build judgement alongside technical skills. Tools change every few weeks, so judgement is the part that lasts.
The second was visibility. In more than one session, the advice to partners was to build personal brands and give individuals in the business a public voice. People buy from people, and that is as true for a Business Central partner as for anyone else.
The third was partnership. Seeing even a small part of the community in the flesh was a reminder of how competitive the market is, and how many chances there are to collaborate. Microsoft 365 partners often hold strong customer relationships but need Business Central expertise to serve them. Working closely with the right partners looks like an underused route to new opportunities for everyone involved.
BC4 works with partners across payments, hardware, POS, ecommerce, data and analytics, inventory and RFID, and recruitment and training. Our partner directory brings them together in one place.
BC4 works with retail and sports organisations on Business Central and LS Retail, from large international events to club and high street operations. If you want to understand where AI can help in your business, and what needs fixing first, get in touch and we will talk it through with you.
Directions EMEA is the main annual conference for the Microsoft Dynamics 365 Business Central community across Europe, the Middle East and Africa. It brings together Microsoft, partners, developers and MVPs. The 2026 event took place in Paris.
It is a Microsoft capability, announced at Directions, that lets users ask questions about their Business Central data in plain English and get answers without opening Business Central. A partner preview was announced for November and December, so timings may still change.
Yes. Business Central licences already include Copilot features such as bank reconciliation, chat, analysing lists, marketing text for items and e-documents to purchase lines. Many customers have not enabled them yet.
For most businesses, not fully. AI can answer quickly, but poor answers usually trace back to messy data, unexplained codes and definitions nobody has agreed. Cleaning up those foundations first gives better results from any AI tool.
It describes a shift where people and AI agents get what they need from an ERP without working in its screens. The value sits in the business logic, trusted data and connected processes underneath, rather than in the interface.
It is an approach where the specification and the tests are written before any code is started. When AI writes much of the code, the spec and tests define what "right" looks like up front, so the result can be checked against them.