If your consulting offer is a polished generic answer, a client should question why they need you. AI has made that kind of output cheap.
The defensible work sits elsewhere: diagnosing the client’s actual situation, obtaining evidence, making trade-offs, implementing safely and remaining accountable when reality refuses to match the slide deck.
Stop selling information
A client can ask an AI tool for a migration plan, Copilot readiness checklist or Power Platform architecture. The answer may sound competent while missing the tenant, politics, constraints and failure history that determine whether it will work.
Your value is not “I know things ChatGPT does not”. Models and search tools keep improving.
Your value is a controlled path from an unclear problem to an evidence-backed decision and a working change.
Five things a useful consultant should own
1. Diagnosis
Turn “we need Copilot” into a specific decision.
Ask who is struggling, which task is failing, what data the task uses, what consequence matters and how success will be observed. Separate the buyer’s requested tool from the underlying problem.
The deliverable is a written problem statement with assumptions and exclusions, not a longer prompt.
2. Evidence
Inspect the client’s real environment with permission. For Microsoft 365 work, that might mean access reports, site ownership, information architecture, licences, support records and representative user journeys.
Use current first-party documentation for product behaviour. Microsoft’s Copilot data readiness guidance, for example, makes access remediation and ongoing governance explicit.
AI can help organise evidence. It does not turn an assumption into proof.
3. Decision design
Give the client choices with consequences:
| Option | What changes | Cost or risk boundary | Proof needed |
|---|---|---|---|
| Do nothing yet | Keep current process | Delay may preserve current pain | Baseline remains stable |
| Pilot | Test a narrow group | Limited evidence, controlled exposure | Defined acceptance cases pass |
| Roll out | Broader operational change | Governance and support load grows | Pilot plus readiness gates pass |
The consultant should recommend one option and explain why. Hiding behind a perfectly balanced options paper is not advice.
4. Implementation and adoption
Create the artefacts that make the decision real: permissions changes, test scripts, operating procedures, owner lists, dashboards, support routes and rollback steps.
Then observe people using the workflow. A system can be technically correct and still fail because staff do not trust it, the exception path is awful or the new step lands on the wrong person.
5. Accountability
Name what you checked, what you did not check and what still needs client approval. Record decisions and hand over ownership.
Never imply that using AI makes you faster, cheaper or more accurate unless you measured that result on the client’s work. A tool demo is not an outcome.
Use AI backstage, with controls
AI can help draft interview questions, compare source material, identify gaps and turn notes into a first structure. The consultant still needs to verify every consequential claim.
For client data, check the exact account and contract. OpenAI says business products and its API do not use organisational inputs or outputs for model training by default, but consumer products, connected apps and client policies can have different terms. Other providers differ too.
Do not paste confidential tenant data into an unapproved account because the output is “only a draft”.
Package proof, not invincibility
Avoid claims that you are irreplaceable. Sell a bounded engagement:
- a permissions and data readiness assessment;
- a pilot with explicit acceptance tests;
- a decision record and remediation backlog;
- a working automation with failure handling;
- a handover with owners and evidence.
The client can inspect those outputs. That makes the value clearer than a promise about expertise.
Boundaries checked on 24 August 2026
This is commercial positioning advice, not evidence that consulting demand will rise or that a particular practice will remain viable. The ILO’s task-level research supports job transformation as a more likely aggregate outcome than full replacement, but it does not predict an individual consultant’s prospects.
For practical help turning AI capability into accountable client work, join the Microsoft Copilot Adopters Space.