Microsoft reported yesterday that Teams now has 250 million monthly active users. Eighteen months ago the number was twenty million. The commentary has already settled into a familiar shape: the collaboration war is over, the bundled product won, everyone else is fighting for the remainder. That reading is broadly correct and almost entirely useless to anyone who has to run the thing. Start with the number itself, because it is being compared carelessly. Two hundred and fifty million is a monthly figure — anyone who opened the application once in thirty days, including the people who opened it because a client sent a meeting link. The company's last daily figure, published in April, was 145 million. Monthly and daily counts measure different behaviours, and the gap between them is precisely the population that has Teams installed and does not work in it. Slack, for its part, stopped publishing daily user numbers two years ago; its disclosed commercial metric is paid customers, and a comparison between a monthly reach figure and a paid-account figure tells you nothing at all. The honest summary is narrower and more interesting. Teams won on distribution, not on product. It is included in a licence that most organisations already buy, it appears in the launcher whether or not anybody asked for it, and the pandemic supplied a use case urgent enough to overcome the usual inertia. Which means the real situation in most companies today is this: you are running an enterprise collaboration platform you never selected, never designed and cannot currently describe.
| Question | What to inspect | Follow-up |
|---|---|---|
| Where is work discussed? | Channels and private chats used for real decisions. | Check whether the chosen record is findable. |
| Which teams still work? | Activity, named owners and business purpose. | Review dormant teams without bypassing retention. |
| Where are files kept? | Shared files, personal stores and permissions. | Test access for intended colleagues. |
| What are meetings costing? | Meeting load and decisions produced. | Review purpose rather than assuming growth is success. |
Qualitative summary of this article's source text, not a measured outcome or performance estimate.
The numbers to measure internally
If the vendor's monthly figure is a marketing artefact, what should a CIO actually track? Four things, none of which appear on the default dashboard. The channel-to-chat ratio. What proportion of messages are posted in channels versus private chats and small group chats? In most estates it is eighty to ninety per cent private. That single ratio determines whether the platform is building organisational memory or a very large collection of unsearchable side conversations. Active teams versus created teams. Count the teams with a post in the last ninety days against the total. A ratio below half means you have a sprawl problem that is currently invisible and will surface as an eDiscovery cost. File location share. How much document collaboration now happens in the platform's underlying storage versus email attachments? Attachments are the signal that the migration is incomplete. Meeting load per head. Hours per person per week, tracked as a trend. Nobody deployed a collaboration platform in order to add six hours of meetings, and yet. Those four numbers describe whether the deployment succeeded. Seats do not.
What goes wrong at this scale
The failures are consistent across organisations and they are all governance failures rather than technical ones. Teams proliferate because creating one is a single click and nobody ever deletes anything, so an estate of four hundred people accumulates nine hundred teams, most with one owner who has since left. Files scatter, because the platform stores documents in a document library that many administrators have never opened and most users cannot find from outside the app. Guest accounts accumulate from projects that ended in 2020. Private chat displaces channels, so the knowledge that was supposed to be captured evaporates into one-to-one threads. And notification defaults train people to treat every message as urgent, which is how a tool designed to reduce email produces a second inbox nobody can declare bankruptcy on. The governance answer is deliberately small: naming conventions, a named owner and deputy for every team, an annual ownership attestation that archives anything unclaimed, an expiry policy for project teams, a quarterly guest access review, and a stated rule about where different kinds of content belong — decisions in channels, drafts in files, records in the system of record. That is weeks of work, not a programme, and it is the difference between a platform and a landfill.
The quiet second act is telephony
The milestone that will matter more to next year's budget is not chat adoption. It is the steady absorption of the desk phone. The platform now carries a full telephony capability: direct routing to your own carrier circuits, calling plans where they are offered, and newer operator interconnect arrangements that let a telecoms provider connect its service to the tenant with less engineering. For organisations with an ageing phone system and a maintenance contract they resent, the business case writes itself — one identity, one client, one directory, and the handset becomes optional. This is also where the bundling strategy becomes clearest. The suite keeps absorbing adjacent categories: meetings, then webinars, then whiteboarding, then employee communications and, this year, an entire employee experience layer. Each absorption converts a line item on somebody's budget into a feature of something already paid for. Buyers should be honest that this reduces their leverage as much as their cost — the default is now decided by the licence, and the default decides the estate.
Practical Guidance for Teams Optimization Strategy
- Measure the four internal numbers — channel-to-chat ratio, active-to-created teams, file location share, meeting hours per head — before buying anything else.
- Assign an owner and a deputy to every team, with an annual attestation that archives whatever nobody claims.
- Set lifecycle rules at creation: naming, purpose, expiry date for project teams, and a defined archive destination.
- Review guest access quarterly and remove accounts from finished engagements; this is the cheapest risk reduction available.
- State where content belongs in one page, and back it with defaults rather than policing.
- Fix notification defaults centrally instead of asking people to fix them individually.
- Model the telephony case properly — carrier interconnect, number porting, emergency calling obligations and per-user add-on licences, not just the removal of the old maintenance contract.
- Check your licence edition against what you actually use; many organisations are paying for a higher tier for two features and not using either.
The Regional Angle
Three issues shape what this platform can actually do for a Gulf organisation, and the first will decide whether the telephony business case survives contact with reality. Voice is regulated here in a way that global deployment guides do not anticipate. Calling plans sold directly by the software vendor are unavailable in most Gulf markets, and the provision of telephony services is licensed, with public network interconnect controlled by the incumbent operators. The practical route is therefore direct routing through a locally licensed carrier's circuits where the operator supports it, with a session border controller either on your premises or provided as a service. That is a real project with a carrier negotiation attached, not a licensing change, and the economics differ sharply by country: in some markets the local trunking is cheap and the case is excellent; in others the interconnect pricing quietly eliminates the saving that justified the migration. Model each country separately, confirm what your operator will actually support before committing to a regional rollout, and check emergency calling and number portability obligations in each jurisdiction rather than assuming the global default applies. The second is tenant geography, and it is worth checking this week because most organisations have never looked. The default storage location for a tenant's data is fixed by the country selected when the tenant was first created — often in 2015 or 2016, often by a reseller, often set to the country of a parent company or simply to whatever appeared first in the list. With regional datacentre capacity now live in the Emirates and further capacity announced across the Gulf, many organisations assume that their data followed the infrastructure. It did not. Default data location is not a setting you toggle; changing it is a migration request with eligibility conditions and a lead time. For any organisation now writing sovereignty language into its policies or answering customer questionnaires about where collaboration data resides, the first action is to read the actual tenant configuration and stop repeating what the sales deck said. The third is tenant fragmentation, which is endemic in regional groups. Entities acquired at different times, in different emirates and different countries, each bought licences through a different partner, each ended up with its own tenant — so a single group operates three or four separate collaboration estates with different domains, different policies, and staff who hold two accounts and use whichever has the meeting. External guest access becomes the workaround, which is how guest lists grow to thousands. Cross-tenant shared channels have been announced but are not yet available in general release, so until they are, the realistic choices are consolidation onto one tenant — a serious, months-long project best justified alongside an identity programme — or an explicit, documented federation design with agreed guest policies, naming and review cycles. What does not work is the current default, which is neither.
The objection worth taking seriously
The objection I hear most from engineering-minded leaders is that optimising a product you get free with the bundle is consultancy looking for work. The platform is good enough, people self-organise, and every governance initiative in the history of collaboration software has added friction, annoyed users and pushed real conversation somewhere ungoverned. Approval workflows for creating a team are the canonical example: they slow work down by a day and send the project to a messaging app instead. That is a fair description of how these programmes usually fail, and anyone proposing a team-creation approval board should be asked to describe the last one that worked. The response is that the choice is not between governance and freedom. It is between cheap guardrails now and expensive archaeology later. The organisations discovering this the hard way are the ones asked to produce three years of chat for a dispute, or to work out which of nine hundred teams contain customer personal data, or to remove a departed employee's access to information nobody can enumerate. None of those problems is solved by a tool purchase; all of them are cheap to prevent with ownership, expiry and review — five rules, applied at creation, automated where possible. The friction objection is valid against bureaucracy and irrelevant against defaults. Set the defaults well and most users will never notice the governance exists, which is the only version that has ever survived contact with a workforce.
Common Questions
Does this mean Slack is finished?
No, and the framing is wrong. Its position now runs through its new owner's customer platform rather than through head-to-head chat competition, and it remains strong where developer workflow and external community collaboration matter. Distribution beat it in the general enterprise market; that is a different statement from irrelevance.
Should we still run two collaboration platforms?
Only with a stated boundary — by function, by audience, or by external partner requirement — and a named owner for each. Two platforms with overlapping purposes produce three places to look for the same conversation.
Is the monthly active user figure meaningful at all?
As a measure of reach and bundling power, yes. As a measure of whether anybody is working differently, no. Use your own four numbers.
What should we expect over the next twelve months?
Expect the platform to push hard into telephony and into frontline and deskless workers, which is where the remaining unlicensed population sits. Expect cross-tenant shared channels to become generally available and to change external collaboration more than any feature shipped this year. Expect the newly acquired competitor to be repositioned around customer and sales workflows rather than chat volume. Expect regulators in financial services to treat collaboration platforms as record-keeping systems in earnest, which will make retention configuration a compliance question rather than a storage one. And expect at least two more product categories to be absorbed into the suite, each converting somebody's standalone subscription into a feature you already own.
Teams Optimization Strategy — we measure how your organisation actually uses the platform, put lifecycle and ownership rules around the sprawl, and test the telephony and tenant decisions against what is genuinely available in your markets.