In July, Microsoft’s Corporate Vice President for Dynamics 365 Customer Experience, Deva Rajamohan, set out the company’s vision for agentic customer experience across Microsoft 365 Copilot and Dynamics 365. The headline was simple: AI should stop being another tool your sellers and service teams have to open, and start working inside the apps they already live in — Outlook, Teams, Copilot and the CRM itself.
Three months on, the products behind that vision are generally available. Sales Agent and Service Agent run inside Microsoft 365 Copilot. Copilot is now embedded directly in Dynamics 365 Sales and Customer Service. And Dynamics 365 plugins for Copilot Cowork extend the same capabilities into team-based work.
For UK mid-market businesses, the question is no longer whether agentic CX is coming. It’s whether your data, processes and people are in a state where an agent can do anything useful with them. That is a far less glamorous question — and it’s the one that decides your return.
What Microsoft actually shipped
Strip away the launch language and there are four concrete things to understand:
| Capability | What it does | Where it runs |
|---|---|---|
| Sales Agent | Pulls account and opportunity context, summarises past meetings, drafts grounded follow-ups, and writes notes and field updates back to the CRM | Copilot, Outlook, Teams, mobile |
| Service Agent | Summarises cases, surfaces knowledge, recommends next best actions, drafts customer responses and updates records | Copilot, Teams, Outlook, Dynamics 365 |
| Microsoft 365 Copilot in Dynamics 365 Sales / Customer Service | Brings the same agents inside the CRM so behaviour is consistent wherever people work | Dynamics 365 |
| Dynamics 365 plugins for Copilot Cowork | Coordinates multi-step, multi-person work — deal prep, escalations, case reviews — as one workstream | Copilot Cowork |
The technical shift underneath matters more than any single feature. These agents are built on Model Context Protocol (MCP) tools — standardised connectors that let an agent not just read your CRM but act on it. At general availability, Service Agent shipped with more than 70 new MCP tools, moving it from answering questions to updating cases, creating activities and closing loops. Microsoft followed up this month with 30 new CRM skills for Copilot and custom agents.
That is the meaningful line between “Copilot” and “agentic”. An assistant tells you what’s in the record. An agent changes the record.
Why this lands now for UK businesses
The case for agentic CX rests on two problems every sales and service leader recognises.
Sellers don’t spend much of their week selling. Salesforce’s State of Sales research found reps spend just 28% of their week actually selling, with the rest absorbed by deal admin, data entry and switching between tools. Two-thirds said they felt overwhelmed by the number of tools they had to use.
UK customers are judging you on effort. The Institute of Customer Service’s July 2026 UK Customer Satisfaction Index puts the all-sector score at 78.3 — up a point on last year — and notes that value, quality and being easy to deal with have all grown in importance for customers over the past six months. “Easy to deal with” is precisely what a service rep with a pre-summarised case and a drafted resolution can deliver.
The numbers behind the pitch
- 54% of UK firms now actively use AI, up from 35% a year earlier (British Chambers of Commerce, 2026)
- ~1 in 10 UK SMEs have gone beyond generic tools to bespoke AI implementations (ISER / University of Essex, 2026)
- 19% of UK small business owners describe themselves as “very confident” using AI day to day (Simply Business SME Insights, 2026)
- 2.6x — how much more likely sales organisations giving reps AI-enabled next best actions are to achieve commercial growth (Gartner, May 2026)
Read those together and a pattern emerges. UK businesses have adopted AI at the generic layer — chat, drafting, summarising. Very few have connected it to the systems that actually run customer relationships. Agentic CX is that connection. It’s also where the confidence gap will bite hardest.
The uncomfortable bit: agents amplify whatever your CRM already is
Microsoft’s own framing is honest on this point: an agent is only as valuable as the data it can reach. It’s worth taking that more literally than the launch post does.
If your opportunity records have stale close dates, the Sales Agent will brief your seller on a deal that isn’t real. If half your service knowledge lives in someone’s inbox rather than in Dataverse or SharePoint, the Service Agent will draft confident answers from the half it can see. If your security roles are loose, an agent that respects existing permissions will faithfully respect the wrong ones.
This is why we’d treat the headline ROI figures with care. The $3.70 return per $1 invested in generative AI that Microsoft cites comes from an IDC study Microsoft sponsored, averaged across large organisations. It may well be achievable — but it isn’t a forecast for your pipeline.
Gartner offers the counterweight. It predicts that over 40% of agentic AI projects will be cancelled by the end of 2027, citing escalating costs, unclear business value and weak risk controls, and estimates that only around 130 of the thousands of vendors claiming agentic capabilities are the real thing. The same firm predicts agentic AI will autonomously resolve 80% of common customer service issues by 2029.
Both predictions can be true at once. The technology is real; most implementations will fail anyway. The difference is almost never the model. It’s scope, data and ownership.
Where Microsoft’s approach genuinely helps the mid-market
To be fair to the launch, a few design decisions make this more viable for a 50- or 300-person business than earlier waves of CRM AI:
- No new interface to adopt. Sellers already live in Outlook and Teams. Putting the agent there removes the single biggest adoption barrier we see in mid-market CRM projects: getting people to open the CRM at all.
- Granular, reversible rollout. Service Agent can be switched on per role, per app module and per queue, and run side by side with existing experiences. That makes a contained pilot genuinely contained.
- Governance in the same stack. With Microsoft Agent 365 now generally available for managing and governing agents, organisations already on Microsoft 365 don’t need to bolt on a separate control plane.
- CRM hygiene as a by-product. When the agent captures meeting takeaways and updates opportunity fields, data quality stops depending on sellers’ goodwill on a Friday afternoon. Over time, the agent improves the very data it relies on.
A practical readiness framework: four questions before you switch anything on
We use a simple test with clients before any agent goes live. If you can’t answer “yes” to all four for a given team, fix that first.
1. Is the licensing actually in place? Service Agent requires Dynamics 365 Customer Service Enterprise or Premium, with a Microsoft 365 Copilot licence unlocking the full integrated experience. Sales Agent has equivalent dependencies on the sales side. Model the per-user cost against a specific team before you model the benefit — not after.
2. Would you trust the records the agent will read? Pick ten live opportunities or open cases at random. Are owners, stages, dates and next steps accurate? Is the relevant knowledge in Dataverse or SharePoint, not in inboxes? If not, a short data-quality sprint will return more than any agent configuration.
3. Do permissions reflect reality? Agents operate within each user’s existing access. Review security roles, business units and SharePoint sharing before rollout — especially in regulated sectors such as financial services and professional services, where UK GDPR and sector rules apply to anything the agent surfaces.
4. Can you measure one number before and after? Choose a single, owned metric per team: time to first response, average handling time, follow-up completion within 24 hours, or forecast accuracy. Baseline it now. If you can’t measure it, you can’t defend the licence spend at renewal.
What good looks like in the first 90 days
| Phase | Focus | Output |
|---|---|---|
| Weeks 1–3 | Data and permissions audit for one team | Cleaned records, reviewed security model, baseline metric |
| Weeks 4–8 | Contained pilot (one sales pod or one service queue) | Usage data, qualitative feedback, first before/after comparison |
| Weeks 9–12 | Decide: scale, adjust or stop | A business case built on your own numbers, not a vendor’s |
The point of this shape is to make “stop” a legitimate outcome. The organisations that end up in Gartner’s 40% are usually those that committed to scale before they had evidence.
The bottom line
Microsoft has done the hard platform work. Agentic capabilities now sit in the tools your teams already use, built on a standard that lets agents act rather than just advise. For UK mid-market firms competing against larger rivals on responsiveness, that’s a real opportunity.
But the agent doesn’t change the fundamentals of good CRM — it exposes them. Clean data, sensible permissions, a clear owner and a measurable outcome were always what separated CRM projects that paid back from those that didn’t. Agentic CX simply raises the stakes and shortens the feedback loop.
Start with one team, one metric and one honest look at your data. Then let the numbers decide.
Thinking about Sales Agent or Service Agent for your business? Kompound helps UK mid-market organisations get Dynamics 365, Power Platform and Copilot ready for agentic AI — from data and permissions audits through to contained pilots with a measurable business case. Talk to us about an agentic CX readiness review.
Sources and further reading
- Microsoft Dynamics 365 Blog — Moving sales and service organizations forward with agentic CX and Microsoft 365 Copilot (July 2026)
- Microsoft Dynamics 365 Blog — Service Agent reaches general availability in Microsoft 365 Copilot (June 2026)
- Microsoft Tech Community — Sales Agent is now generally available
- Microsoft Dynamics 365 Blog — 30 new CRM skills for Microsoft Copilot (October 2026)
- Microsoft Learn — Use Copilot Cowork sales skills in Dynamics 365 Sales
- The Institute of Customer Service — UKCSI July 2026
- British Chambers of Commerce / Atos — AI adoption among UK SMEs climbs to 54% (March 2026)
- ISER, University of Essex — AI Adoption and Workforce Change in SMEs (March 2026)
- Simply Business — SME Insights Report 2026
- Salesforce — State of Sales research
- Gartner — AI-enabled next best actions and commercial growth (May 2026)
- Gartner — Agentic AI will resolve 80% of common customer service issues by 2029 (March 2025)
- CRN — Gartner: 40% of agentic AI projects could be cancelled by 2027
- IDC InfoBrief (sponsored by Microsoft) — The Business Opportunity of AI (November 2024)
