Scott Adams Agent Sales GBB
Microsoft

Agent Sales Global Black Belt · Microsoft

Change the work.
Prove it can run.

I advise enterprise leaders on where agents can change a business process — not just add another interface. Then I lead the hands-on POC and design the path to governed production at scale.

Brief bio

I'm Scott Adams. Across 30+ years as a software engineer, product leader, founder and enterprise technology leader, I have helped turn emerging technology into operating change people can trust. Today I bring that experience to agentic AI at Microsoft.

I start with your outcome and your process, then choose the right surface across Microsoft 365 Copilot, Copilot Studio, Foundry, GitHub and MCP. The goal is a credible proof in your environment — with the architecture, governance and scale plan to take it further.

Explore a working POC See the reference builds
30+
Years in enterprise technology
Inventor
US patent · data architecture
Engineer
1st-class BEng · software engineering
Author
Microsoft Press co-author
Scott Adams, Agent Sales Global Black Belt at Microsoft
What customers bring me in to do Advise. Prove. Scale.

From business outcome to a governed, production-ready path.

Transformation lab

Don't just read about it. Run it.

Start with an event-triggered Digital Customer Success Manager: a source-aligned, redacted replay of the real seven-stage engine with every external call and side effect disabled. The other three scenarios connect to public demo MCP endpoints on Azure, invoke fixed read-only tools, and host the actual Apps SDK widgets they return.

Demonstration data only. The CSM operational dashboards remain identity-gated and are never embedded here; live widgets run in isolated sandboxes and cannot execute follow-up tools.

Choose a transformation 1 SAFE SIM · 3 LIVE
Ready to replay a business event
local · redacted · side-effect free
See what runs

event → Agent ID → seven-stage engine → deterministic route → supervised outcome

The replay follows the repository's real stages and rule IDs using synthetic inputs. It makes no external or API requests, writes no data and cannot deliver customer outreach.

The interaction surface

Widgets turn answers into work.

Rich MCP Apps let people inspect, navigate and act without forcing every task through a text transcript.

Interactive surface

Widget preview

Working reference builds

The repositories prove the method.

Each reference build starts with a different operating problem. The technology matters because it makes the new way of working possible — not because the architecture diagram is the outcome. Together they show the method: qualify the process, choose the best surface, prove it hands-on, and design the controls and operating model needed to scale.

Flagship transformation build
Product signal received Autonomous
Deterministic decision Human judgement required Strategic account · complex topic · high influence
Manager1 review item SponsorFleet visible GovernanceEvidence captured

01 · Autonomous customer success

A customer signal becomes governed work — without waiting for a prompt.

csm-aiteammate

A Digital Customer Success Manager starts when a usage, release or renewal event lands. It acts with its own Agent 365 identity, builds context, applies auditable routing rules, drafts from approved content and either handles the work or asks the human CSM for judgement.

7
Stage autonomous journey
16
Business MCP tools
3
Supervision surfaces
0
Human turns to start

Transformation pattern: replace periodic queue management with continuous, event-driven customer success — while preserving manager judgement and sponsor control.

Open the case study → GitHub ↗ Run the safe replay ↑
ESS-MCP people leader dashboard Enterprise reference system

02 · Enterprise services

One shared MCP capability layer. Multiple ways of working.

ess-mcp

This reference build demonstrates how employees and managers could complete cross-system tasks with less navigation. Shared MCP servers support a Copilot declarative agent, MCP Apps in compatible hosts, five Copilot Cowork plugins and an autonomous demo integrated with Microsoft Agent 365.

7
Enterprise MCP servers
204
MCP tools
76
Interactive widgets
4
Worked implementation patterns

Transformation pattern: replace duplicated integration work with a shared MCP capability layer, then apply identity, policy and operations end to end for each host and runtime.

Open the case study → GitHub ↗ Try the mocked Coupa flow ↑
Retail daily operations dashboard Retail stock health screen Retail team rota screen Live workflow available

03 · Frontline operations

A shift manager gets a thinking partner.

retail-mcp

Weather, travel, rota, stock, training, policy and customer signals become one continuous operational conversation — designed for people who do not sit at a desk.

19
Tool modules
17
Widgets
15
Demo scenes

Transformation pattern: move from dashboards that report work to an agent that helps run the shift.

Open the case study → GitHub ↗ Run it live ↑
L2Q lead-to-quote pipeline widget Live workflow available

04 · Revenue operations

Salesforce becomes work, not a destination.

L2Q

A regulated fintech sales team moves from CRM lookups to account briefs, meeting preparation, proposals, pipeline packs and follow-up work across Copilot and Cowork.

54
Live MCP tools
8
Branded widgets
2
Agent surfaces

Transformation pattern: use one MCP layer for fast transactional chat and deeper artefact-producing work.

Open the case study → GitHub ↗ Run it live ↑
Digital Lawyer activity and billing dashboard Governed digital worker

05 · Regulated digital labour

An AI colleague with identity, authority and supervision.

digital-lawyer

Alex Morgan reviews contracts clause by clause, marks up Word documents, records risk decisions, escalates beyond delegated authority and bills time — with an Entra-backed identity and audit trail.

13
Step review workflow
18
Step email workflow
5
Domain MCP servers
1
Named agent identity

Transformation pattern: manage capable agents as members of the workforce, not anonymous automations.

Open the case study → GitHub ↗

How I work with customers

From a valuable process to production-scale change.

Outcome first. Best surface. Real proof. Governed scale.

A POC, not a theatre demo Real process. Real systems. A named path to production.

The proof is time-boxed, grounded in representative customer data and built with production patterns. Success means the business outcome, security boundary, operating model and next scale decision are all clearer than when we started.

01

Qualify the outcome

Find the process where context, judgement and action create meaningful value. Define the trigger, the changed way of working and evidence that would make the case credible.

Output · outcome and trigger
02

Scope the real process

Map the data, systems, people, decisions and controls involved. Distinguish human-led assistance from event-triggered process transformation so each is designed appropriately.

Output · POC scope and exit criteria
03

Build on the best surface

Lead the POC hands-on using the right mix of Microsoft 365 Copilot, Copilot Studio, Foundry, GitHub, MCP and custom code — never forcing the problem into one product.

Output · working, instrumented POC
04

Prove it can run

Test the outcome and the production realities: identity, permissions, grounding, human judgement, observability, cost, safety and operational ownership.

Output · evidence and production decision
05

Convert, govern and scale

Turn the proof into reusable architecture, governance, runbooks and an adoption plan so engineering teams and delivery partners can scale the pattern with confidence.

Output · governed scale plan

Experience customers can rely on

Engineering depth. Product judgement. Transformation leadership.

I have spent more than three decades turning emerging enterprise technology into operating change people can trust. That path has taken me from hands-on software engineering and BI product leadership to founding a company, leading technical teams and advising enterprise and public-sector leaders on consequential technology decisions.

Customers bring me the difficult, cross-system process — not a product shopping list. I work at both altitudes: the executive case for change, and the architecture, code and controls that prove the new operating model can actually run. The advice is platform-led; the proof is hands-on; the destination is governed production, not a successful demo.

1st-class BEng · Software Engineering Microsoft Press co-author US patent · data architecture Engineer · founder · product leader

Viewpoints

The ideas behind the builds.

All articles on LinkedIn ↗
Interactive MCP App people-leader dashboard
Jul 2026 · Interaction design

MCP Apps: when enterprise agents stop feeling like chat

Why tool calling is not the user experience—and what a serious POC should prove about usability, economics and governance.

Read →
Digital labour article cover
Feb 2026 · Identity and governance

Digital labour, made real

Why capable agents need identity, delegated authority, supervision and a place in the operating model.

Read →
Microsoft Frontier Agent Platform diagram
Jan 2026 · Enterprise platform

Microsoft Frontier Agent Platform

The control plane, tools and open protocols required when an organisation moves from assistants to an agent workforce.

Read →
Chat and Cowork article cover
2026 · Interaction model

Chat, Cowork, and the plugins that do the work

One MCP layer can power quick transactions and deep work that produces finished dashboards, decks and reports.

Read →

A useful first conversation

Which business process is worth transforming — and what would prove it?

Bring the business outcome, the process, the systems and the hard constraints. I will help you qualify the opportunity, choose the right platform approach, shape a credible POC and define the path to scale and govern it if the proof succeeds.

Start a conversation on LinkedIn Explore all GitHub repositories