Honest fit criteria for 50–2,500 person companies considering Claude Cowork. Strong-fit signals, weak-fit signals, the mid-market trap, and the wait-and-watch zone.
TL;DR. Cowork fits 50–2,500 person companies whose work centres on Office files, PDFs, and email — usually finance, ops, marketing, legal, or HR. It does not fit teams that live entirely in cloud-only platforms, regulated environments without an Anthropic procurement path, or organisations without an owner for rollout. The biggest reason Cowork deployments fail is not technology; it is mid-market companies trying to roll it out using freelancer-shaped guides.
Three lenses on the same question.
Company shape. US- or UK-headquartered, 50–2,500 FTEs, document-heavy functions, mature enough to have shared drives and a security team but small enough that a Big-Four transformation deck is overkill. If you are trying to decide whether to bring in McKinsey for an AI strategy, you are too big for our advice; if you are wondering whether the founder should be doing this on their personal laptop, you are too small.
Function shape. Finance, RevOps, Marketing Ops, Legal, Operations, Customer Ops, HR Ops — wherever someone is paid to wrangle Office files, PDFs, and email. The roles whose week is mostly meetings and decisions get less from Cowork; the roles whose week is mostly artifacts get the most.
Buyer shape. Either a function head with budget authority for software up to roughly $50k/year and a working relationship with IT, or a CFO/COO sponsoring a horizontal AI initiative. Cowork rarely lands well when the buyer is the IT department alone — IT can pick the tool and own the security review, but a function leader has to own the workflow change, or it does not stick.
Eight things that, when present, predict Cowork will earn its keep within a quarter:
The presence of three or more of these is enough to pilot. Five or more, and the question shifts from whether to which function ships value first.
Five things that should give you pause:
If two or more of these apply to you, defer rather than push through.
Three honest pitfalls we see at companies that should have succeeded:
The "every team gets a licence" buy. Without role mapping, half your seats sit idle and the other half are using Chat when they should be in Cowork. This is the single most expensive mistake at mid-market, because it is invisible — the dashboard shows seats provisioned, not value extracted. See Cowork vs Chat vs Code for the per-role mapping we use.
Skipping shadow sessions. Buying the tool before watching how the work happens means you build skills nobody asked for. Two days of observation prevents two months of building skills your team will not adopt. The full method is on Observe → Co-design → Ship.
No folder policy. Cowork edits real files. Without a workspace pattern (Section 02), one user can rearrange a shared drive and ship chaos to fifty people. The fix is small — a half-day of folder convention work — but it has to happen before, not after.
The pages on cowork.tinkso.com sell us when we should be helping. So here is the honest section: when do we tell prospects to defer?
If your security review for new SaaS takes four months or more. Start the Anthropic security review now — it will pay back later. But do not promise leadership a Q3 rollout if procurement is going to consume Q3 by itself.
If your team is mid-ERP migration (or mid-anything-foundational). Cowork on top of a moving foundation amplifies confusion. The right time is six months after stabilisation, not in parallel.
If you have no shared definition of where files live. Cowork accelerates whatever pattern you already have — including bad ones. If your shared drive is genuinely chaotic, fix that first. A simple folder cleanup project (which Cowork can help with, ironically) is the right precursor.
Concrete metrics, not vague "productivity":
These numbers are not aspirational — they are the post-engagement medians we see when the rollout is run properly. Full instrumentation guidance is in Measuring ROI.
If you fit the strong-fit profile, the question is not "should we use Cowork?" It is which function ships value first, and who owns it for the next six weeks?
We will help with the second question. The first is on us to validate in week one of an engagement — by watching the work, not by reading your org chart.
A ten-minute internal test before you commit:
That is enough signal to start a two-week shadow. It is not enough to commit to a year of seat licences. Use one to earn the right to do the other.
Book a 30-minute call. We'll ask where you are, what your team needs, and which systems Cowork should touch.