Most Copilot rollouts at 100-person companies fail for the same reason: someone bought licences before anyone defined the outcome. Six months in, usage has settled at 20 percent, no one can name a specific hour saved, and the renewal conversation gets awkward.
This is the playbook we use to avoid that.
What a 100-person company actually looks like
For this guide, "100-person company" means:
- Between 75 and 150 employees
- Mostly on Microsoft 365 (E3 or Business Premium)
- SharePoint and OneDrive used but not rigorously governed
- A mix of front-office (sales, ops, finance) and back-office (HR, IT, operations)
- No dedicated AI lead — usually a Director of IT or Head of Operations owning the rollout on top of an existing job
At this size you cannot afford a failed pilot, and you cannot afford a consultancy that treats you like an enterprise. You need tight scope, a clear outcome, and proof before you scale.
The three things to decide before you buy a single licence
- Which workflows will Copilot touch first? Not "everyone gets Copilot." Which named workflows, for which teams, with which current hours spent.
- What is the permissions risk? Copilot surfaces anything a user already has access to. If your SharePoint has not been audited in two years, you have a governance problem, not a Copilot problem.
- Who owns the pilot? Not a committee. One person with time allocated and authority to kill or scale the rollout.
If you cannot answer those three, do a one-week readiness review before you buy licences. It is cheaper than a year of shelfware.
The 30-day pilot that actually produces signal
A 100-person company should not roll Copilot out to everyone at once. You will not know what worked.
Pilot group: 10–15 users, hand-picked across at least three functions (operations, finance, a sales or customer-facing team).
Duration: 30 days, not 90. Any longer and the pilot fatigue takes over.
Workflow scope (pick three, no more):
- Meeting summarisation and follow-ups (Teams + Outlook)
- First-draft document generation for a recurring deliverable (board packs, status reports, proposals)
- Inbox triage and reply drafting for one specific queue
Measurement (the part most skip):
- Baseline the current hours spent on the three workflows before you turn Copilot on. Time-tracked or self-reported, but recorded.
- Collect weekly qualitative feedback from pilot users, no more than ten minutes each.
- At day 30, re-measure the same workflows. If the hours saved, drafting quality, or meeting follow-up coverage did not improve materially, Copilot is not the bottleneck — process is.
What goes wrong (and what to fix before it does)
Permissions leakage. Run a tenant-level sharing audit before pilot. Look for externally shared folders, overly broad "Everyone except external" permissions on sensitive SharePoint sites, and inactive guest accounts. Fix the top ten hotspots before Copilot reads them.
Missing Copilot context. Copilot is only as good as the content it can see. For the three pilot workflows, document exactly which SharePoint sites, which mailboxes, and which documents are in scope. Surface them in sensible places. If the pilot team has to hunt for context, they will stop using it.
No shared prompt library. Pilot users default to whatever they learned from YouTube. Build a shared prompt library of ten to twenty prompts that produce the three target outcomes. Store it somewhere everyone can see. This is the single highest-leverage change you can make.
Expectation drift. Copilot is not an agent. It will not do work unattended. Frame it as "first draft faster, review still required." Teams that expect Copilot to replace a step are disappointed. Teams that expect it to accelerate a step are delighted.
The real cost at 100 seats
As of the current published pricing, Copilot for M365 sits in the $30/user/month range for organisations on E3 or Business Premium. At 100 seats that is meaningful: roughly $36K a year in licences alone, before rollout time. A mid-market company should model:
- Licences: $30/user/month × seats committed
- Rollout time: 10–20 hours per week of an internal owner for the first 90 days
- Training and prompt library: one or two focused sessions plus maintained docs
- Governance remediation: a one-time SharePoint and permissions cleanup project
A disciplined rollout at 100 seats should return that inside year one if you scope it to real workflows. Without scope, it will not.
What success looks like at day 90
- Pilot ROI is documented with specific before/after numbers on the three target workflows
- A written go/no-go decision on scaling from the pilot cohort to the full organisation
- A governance baseline: permissions cleanup complete, sensitivity labels in use, a data handling policy published
- A shared prompt library with owners for each prompt
- Copilot usage concentrated on the workflows where it measurably helps, not broad "use it anywhere" sprawl
Anything else is a vibe check dressed up as a rollout.
When Copilot is the wrong answer
At 100 people, Copilot is often the right place to start — but not always. If your team is on Google Workspace, Copilot is a poor fit. If your biggest pain is a custom workflow that requires reasoning on external documents, a custom agent will outperform Copilot at half the licence spend. If your SharePoint is ungoverned, fix that first.
If you are not sure, take the AI Workflow Automation Scorecard. If Copilot is the right first move, it will surface clearly. If something else fits better — a workflow assessment, or a targeted agent — that will surface too.
What we recommend
Run a one-week readiness review. Decide the three pilot workflows. Ship a 30-day pilot with measurement. Scale only the parts that produced signal. Resist the temptation to do "Copilot for everyone" on day one.
If you want the pilot designed, measured, and handed off with documented ROI, the Microsoft Copilot Pilot Sprint runs exactly this playbook end-to-end.