Everyone announced AI in FM this year. Ours was already at work.

A good year for an old problem

Something worth acknowledging happened this year: the facilities management software industry stopped arguing about whether the back office should run itself.

For a long time, the people doing the work knew something the systems did not reflect. The job management platforms recorded what happened. The coordination that made things happen, the chasing, the checking, the matching, the reminding, lived somewhere else: in inboxes, in spreadsheets, and in the heads of the most experienced people on the team. When those people went on holiday, the operation got slower. When they left, the operation got worse.

Now the largest platforms in the sector are saying the same thing, in their own language, on their own roadmaps. Agents that answer the phone. Agents that chase the paperwork. Agents that validate the invoice. The diagnosis we have been making since day one is becoming the consensus view of the whole industry.

We welcome that, plainly and without hedging. When an entire sector converges on the truth, the people who work in it benefit. Competition on this problem will make every product better, ours included.

But consensus on the diagnosis is not consensus on the treatment, and not everyone arriving at it is arriving the same way.

Two ways to arrive

First, a word about who is writing this. TYTEN is a young company, and in this category that is precisely what makes us the incumbent: our AI workforce was chasing real job sheets, validating real supplier invoices, and answering real service desk calls while the market was still choosing its keynote slides. The work orders are counted in the tens of thousands, and the results are published with clients' names on them. When the arrivals of 2026 describe what agents will do for FM, they are describing our production system in the future tense.

The vendors arriving now split into two camps, and it is worth being precise about them, because one deserves a welcome and the other deserves a caveat.

The adopters use the new words and mean them the ordinary way. Established platforms adding AI features to products they already run: a summary here, a triage model there, an assistant in the corner of the screen. This is genuinely good. It is how new technology reaches everyone, and their customers will be better off for it. But an AI feature inside a job management system is not the same product as an AI workforce built FM-first to run the back office end to end. Both can be worth buying. They should not be confused, and they should not be priced alike.

The imitators take the other route. Rather than evolving an original product of their own, they adopt the original's category, its vocabulary, its module list, occasionally its sentences, and present the resemblance as equivalence. The resemblance is skin-deep by necessity, and the place the difference surfaces is the one place a website cannot hide it: the terms. The original signs you for a month at a time, because it expects to earn the next one. The imitation asks for five years, because it expects you to notice.

Here is the generous part, because fair is fair: some of what the imitators are building is beginning to do genuinely useful work. An agent that drafts a schedule or chases a visit confirmation is real admin lifted off a real person's desk, whoever shipped it, and the industry is better for every hour of it. The point is not that the copy does nothing. The point is that you should not sign original-length commitments, at original-confidence prices, for it.

Enterprise software has met both camps before. Several times. The history is worth the detour, because it says exactly how this ends.

Every platform shift arrives twice

Each generation of business software has had its moment like this one: a genuinely new capability appears, and the question becomes who gets to set the terms of it.

In 1984, computing lived in the mainframe: centralised, standardised, sold on terms the vendor controlled. Apple's famous Super Bowl advertisement did not list a single feature of the Macintosh. It attacked conformity itself, and offered the individual a machine of their own.[1] The product was personal computing. The message was choice.

In 2000, business applications meant installed enterprise software: multi-year licences, armies of implementation consultants, deployments measured in years, and annual maintenance fees on top. A tiny company called Salesforce staged a mock protest outside the industry leader's own user conference, declaring "the end of software."[2] The target was never really the competitor. It was the business model: pay enormously up front, wait a year to go live, then keep paying whether it worked or not. A thousand organisations signed up within two weeks.

The pattern repeated with the maintenance-fee era that followed, when the giants of on-premise software turned annual support charges of twenty percent and more into their most profitable product, used licence audits as leverage at renewal, and let multi-year contracts renew on autopilot.[3] An entire industry of third-party support and cloud alternatives grew up specifically to unpick it. And it repeated again in workplace software, when a young messaging company greeted the largest software vendor on earth entering its market with a full-page letter in the New York Times, echoing the advertisement Apple had once run to welcome IBM to personal computing: confident, gracious, and completely unwilling to compete on the incumbent's terms.[4]

The lesson across forty years is consistent, and it is not that incumbents are villains. Their behaviour is rational: when a new capability threatens an existing revenue model, the safest move is to deliver the new capability inside the old model. Longer contracts. Deeper integration into the existing estate. Pricing decided after the customer is committed. The lesson is that industries eventually move to whichever vendor aligns the new capability with the customer, and history is unkind to the wrapper.

FM software is now at exactly that fork. The capability, an AI workforce for the back office, is real. The open question is whether it arrives on terms built for the operator, or terms built for the vendor's installed base.

Imitation is the sincerest form of roadmap

There is one more chapter in the pattern, and every challenger learns it eventually: before the incumbent ships the product, it ships the language.

The sequence is dependable enough to set a watch by. A smaller firm spends years in the field proving that something the industry accepted as inevitable is in fact automatable, and publishes the results. Then the incumbent's keynote adopts the diagnosis. Then the incumbent's website begins to rhyme with the challenger's website: the same pains, the same promises, occasionally the same sentences lightly reworded. And then, some quarters later, a programme is announced, in which the words are available immediately and the working system is scheduled for a future wave.

Technology history treats this as routine. When a photo-sharing app invented disappearing stories, the feature reappeared, name included, across every product its much larger rival owned; the app's founder replied that just because Yahoo had a search box did not make it Google.[5] Apple absorbed so many ideas from smaller developers that the developers coined a verb for it, "sherlocked."[6] The messaging company that received that famous full-page welcome letter spent the following years watching its vocabulary, its interface, and eventually its entire category description reappear in its competitor's launch materials.

Operators should read this correctly, because we certainly do. It is not an outrage. It is confirmation. Imitation is how an incumbent announces, in its own way, that the challenger was right, and the challenger should accept the compliment gracefully. First they ignore you, then the brochure starts quoting you.

But there is a reliable way to tell the original from the tribute act, and it is not the wording, which is free to borrow. It is the terms. The published, named results. The go-live measured in weeks. The price in writing. The month's notice to leave. Those never survive the copying process, because they are structural, and structure is precisely what the old model cannot copy without giving itself up. A challenger's website can be paraphrased in an afternoon. A challenger's contract cannot, and the contract is the part the customer actually lives with.

When the brochure rhymes and the terms do not, you have learned everything you need to know.

What the automation is for

Before the terms, the purpose. It is easy to talk about AI in FM as if efficiency were the destination. It is not. Efficiency is the vehicle. The destinations are older and more important than any technology cycle:

Buildings that are safe and compliant, all the time. Not compliant at audit. Not compliant on the dashboard refresh. Compliant in the sense that the certificate exists, the remedial was raised, and the evidence is on file before anyone asks for it.

A supply chain where good work gets paid promptly. The subcontractor who did the job well, submitted the sheet, and invoiced correctly should not wait on a month-end batch or a back-office backlog. Cash flow is not an accounting detail. For thousands of small contracting firms, it is survival.

Back-office careers that are worth having. The people who run FM operations are skilled at judgement: handling the difficult client, the ambiguous callout, the supplier relationship that needs a conversation. They are not skilled at retyping PDF fields into a CAFM, because nobody is, and nobody should have to be.

Clients who receive evidence, not effort. The measure of a service desk is not how hard it worked. It is whether the client can see, at any moment, that their portfolio is under control.

Any AI initiative in this industry, from any vendor, should be measured against those four outcomes. Everything else is packaging.

Five ground truths

With the history and the purpose in view, here is what we believe operators are entitled to expect from anyone selling them automation. These are not features. They are terms of trade for the agentic era, and we invite readers to apply every one of them to us as strictly as to anyone else.

1. The operator chooses, and keeps choosing.

This is the ground truth the last forty years were about, and it has two halves.

The first is the stack. Automation that only works if you already run one vendor's platform is not a service to the industry. It is a feature of that platform, and the real product being sold is the wall around it. An operator's choice of CAFM should be driven by their contracts, their clients, and their delivery model. It should never be held in place by the fear of losing access to intelligence.

The second is the contract. The five-year term did not survive from the mainframe era because customers loved it. It survived because it converts a vendor's future performance into a prepaid certainty, for the vendor. A long lock-in is not a partnership. It is an insurance policy against the customer noticing. A vendor confident in its product earns the renewal continuously, and a vendor that needs a five-year term to keep you is telling you what it expects year two to feel like.

And the term itself is only the frame. Enterprise software spent decades perfecting the quiet mechanics that live inside it, and operators who have run a large CAFM or ERP estate will recognise every one:

The change invoice. Once the term is signed, change becomes the product. Every new field, report, workflow or user role is scoped as a professional services engagement, quoted in consultant day rates, and scheduled to the vendor's calendar. The system you bought to move faster now bills you for moving at all.

The reset. Add a module or extend a licence mid-term and the amendment quietly restarts the full term, so the five years begin again from the date of the addition. Grow with the product on these terms and the contract becomes perpetual by instalments: a treadmill dressed as a partnership. The customers who use the platform most are, by design, the ones who can never reach the end of it.

The ratchet. Licence counts and volumes can be revised upward at any moment, and downward only at renewal, if at all. The industry even has a gentle name for the annual exercise of paying for the growth, the "true-up."[7] There is no true-down. Shrink your operation, lose a contract, restructure a team, and you keep paying for the headcount you no longer have.

The escalator. An annual uplift of inflation plus a margin, compounding inside a term you cannot exit. The price you negotiated is the price of year one only; the term guarantees you will be there to pay the others.

The window. The contract renews itself unless notice is served inside a period buried in a schedule, typically a quarter of a year before the date anyone remembers.[8] Miss it by a day and the term begins again. Entire software businesses now exist purely to warn customers about other software businesses' renewal clauses, which tells you everything about whose interest the clause serves.

The exit bill. Your own data, returned slowly, in formats chosen for inconvenience, priced by the export. The final invoice arrives for the privilege of leaving.

None of these appear in the demo. All of them appear in year two. And every one of them shares a single design principle: the value flows from the product, but the profit flows from the term.

The test: if this stops earning its keep, what does leaving cost me, in money, in notice period, and in getting my own data back? What does every change cost after signature? And if I add a module in year two, does the clock restart?

2. Live contracts are not a laboratory.

A new ritual is spreading through enterprise software: the invitation-only early access programme. You run unfinished automation across your live contracts. You attend the cadence calls, supply the feedback, star in the case study. The pricing arrives later, once you are embedded.

Now look at who carries what. The missed SLA is yours. The bruised client relationship is yours. The product improvements belong to the vendor. Stripped of the branding, it is free product development, performed by the customer, at the customer's risk.

Evidence should come before enrolment, not from it. A vendor asking to touch your contracts should arrive with published results from a named operation: numbers achieved, not projected.

The test: show me what this did for an operation like mine, in public, before it touches my contracts.

3. A price is a commitment, and it proves itself in weeks.

A price does three things. It lets an operations director put a number in a budget. It lets a board compare that number against headcount, offshoring, and doing nothing. And it commits the vendor: this is what we believe the value is worth. Automation offered free now, with pricing to follow once the operator is embedded, defers all three. By the time the number arrives, the switching cost has already been built. Enterprise software has run that sequence before, and the number that arrives after commitment is rarely the number that would have been quoted before it.

Timescale is the other half of the same commitment. FM is a thin-margin, high-pressure industry where mobilisations are measured in weeks and client patience in missed SLAs. An automation programme whose own literature talks about first signals in a quarter and full effect in half a year is a roadmap asking to be funded on faith. A product is something that is doing the work, measurably, within the first month.

The test: what does this cost, today, in writing, and what exactly is live and producing output in week four?

4. Compliance does not belong on a future release.

Statutory compliance is a present-tense obligation. The certificate is due when it is due. The remedial from a failed inspection needs raising the day the report lands, not when the relevant module reaches general availability.

Any automation strategy for FM that treats compliance as a later phase has the priorities inverted. Compliance chasing, evidence checking, and remedial detection are not the advanced course. They are the entry requirement, because they are where the industry's real risk lives.

The test: is compliance automation running on day one, or is it on a wave of the roadmap?

5. Everyone should be able to tell the people from the software.

There is a fashion for dressing automation up as colleagues: giving each function a human first name, a face, a job title borrowed from the org chart. The intent is warmth. The effect, in a working operation, is confusion.

When "the helpdesk" is both a person and a persona, accountability blurs at exactly the moment it needs to be sharp. Who confirmed that visit, your coordinator or the software? Which one does the client think they emailed? And when the audit question comes, "who made this decision", the answer should never require remembering which human name a vendor assigned to which module.

Software should be named for what it does, so that anyone in the operation, and any client on the other end of it, can tell in one glance what is a person exercising judgement and what is a system executing a rule. People deserve names. Functions deserve labels. An industry built on compliance evidence should not accept ambiguity about who, or what, did the work.

The test: can everyone in my operation, and my clients, always tell whether they are dealing with a person or a system?

Where we stand, held to the same tests

It would be evasive to write this and not answer our own questions, so, briefly and factually:

TYTEN runs on the systems operators already have. Twelve CAFM and field service platforms are live today, alongside the major accounting packages and ordinary mailboxes. No migration, no new portals for teams or contractors, and your data stays in your systems, because that is where we put it. Commercially, the same principle holds: each module is contracted on its own, priced on the volume you actually use, with a one-month break clause. Adding a module starts nothing over; it is simply another module, on the same terms. Configuring the system to your processes is part of the service, not a change order. If TYTEN stops earning its keep, leaving costs you a month's notice, and nothing else.

Our results are published, with the client's name on them. Penguin FM, coordinating 250 subcontractors, grew work order volumes fivefold while helpdesk capacity increased by 40 percent and the backlog cleared entirely.

Go-live is four weeks for the first module, priced before the engagement starts, and every module works independently, so an operator can begin with the workflow that hurts most.

Compliance is not a future wave. Certificate chasing, document validation, and remedial detection were among the first things we built, because they were the first things the industry needed.

And our modules carry the names of the work they do. Job Sheet Chasing chases job sheets. Nobody in a client's operation has ever had to ask what it is, who it is, or whether it is a person.

We publish these not as a victory lap but as an invitation: hold us to the five tests, permanently, and hold everyone else to them too.

Welcome, seriously

When IBM entered personal computing in 1981, Apple took out a full-page advertisement that began "Welcome, IBM. Seriously."[4] Decades later, a messaging startup did the same on the day the world's largest software company launched a competitor to it. The gesture was not naivety. It was confidence, and a statement about whose terms the contest would be fought on.

So, to the platforms now arriving at the problem we work on every day: welcome, seriously. The FM back office has carried this industry manually for long enough, and it will take more than one company to change that. And if, on the way in, some of our sentences have proven useful, consider them a gift. The results they describe are harder to pack.

The future every serious vendor is now pointed at, compliance as a continuous state, headcount that stops scaling with volume, skilled people doing judgement instead of data entry, is worth pointing at. But direction is not destination. The industry will get the version of that future it negotiates for. If operators accept automation that locks them to a single stack, holds them for five years, tests itself on their live contracts, defers its pricing, and schedules compliance for later, that is the version that will harden into the standard.

Ask the five questions. Ask them of every vendor with an agent to sell. Ask them of us first.

Sources & further reading

1. Apple's "1984" advertisement. Aired once nationally, during Super Bowl XVIII; directed by Ridley Scott; positioned the dominant computing model as conformity and the Macintosh as choice.
     Wikipedia: 1984 (advertisement)
2. Salesforce's "End of Software" launch. Hired actors staged a mock protest outside the Siebel user conference in San Francisco, February 2000; roughly a thousand organisations signed up within two weeks.
     PRovoke Media: the launch of salesforce.com, SMEI: early guerrilla marketing tactics of salesforce.com
3. The maintenance-fee era. Annual support at 20–22% of licence value, audit leverage at renewal, and the third-party support industry that grew up to counter it.
     Spinnaker Support: counter-measures for enterprise lock-in
4. "Welcome" letters. Apple's 1981 "Welcome, IBM. Seriously." advertisement and Slack's 2016 full-page "Dear Microsoft" letter on the day Teams launched.
     Slack: Dear Microsoft, Adweek: Slack's letter echoes Apple's famous ad
5. "Just because Yahoo has a search box." Snap's founder on feature copying, Q1 2017 earnings call.
     CNBC: Snap CEO on copying, May 2017
6. "Sherlocked." The term the developer community coined when platform vendors absorbed smaller developers' products.
     Wikipedia: Sherlock (software)
7. The one-way ratchet. Enterprise agreements commonly allow licence counts to rise mid-term ("true-up") but not fall until renewal, if at all.
     Redress Compliance: the enterprise agreement true-up
8. Auto-renewal notice windows. Enterprise contracts typically require 90 days' notice ahead of auto-renewal; a category of software now exists to track other software's renewal clauses.
     ExpiryEdge: the notice window is the deadline

Historical examples describe patterns across the enterprise software industry over four decades. No reference to any current FM software vendor is made or implied; the five tests are offered for the reader to apply to every vendor in the category, TYTEN included.

---