Your unofficial record is the one that gets read out
Every business we work in has two records. There is the official one, in the system somebody paid for, and there is the real one, in a group chat on people's phones. The second is faster, more complete and more candid. It is also, increasingly, the one a court reads.
Why did WhatsApp win, and who is actually in the groups?
It is tempting to read group chat as a discipline problem. It is not, and treating it as one is why so many policies fail.
WhatsApp won on friction. There is no licence to buy and no seat to allocate. There is no login at seven in the morning when a delivery is already on its way, no training session, no helpdesk ticket when somebody's password expires. A photograph takes two taps and arrives at a size that works on a weak signal. A voice note works with gloves on. The app is already installed, and it already holds the phone number of everyone involved.
That last point is the one enterprise software keeps losing on. A platform gets rolled out to your own staff. It rarely reaches the subcontractor's foreman, the plant hire desk, the supplier's driver, the client's facilities manager who wants to see something before it is closed up. Those people are central to the work and peripheral to the licence count, so they stay in the chat. Per seat pricing is not an incidental detail here. It is the mechanism that guarantees the most operationally involved people are outside the system of record.
So the traffic layers itself. Directors use it between themselves for speed. Managers use it to reach people who are not at a desk. Supervisors use it downward to crews and outward to suppliers. Crews use it between themselves. Each layer is moving information that used to be a letter, an instruction or a minuted decision.
What is passing through those groups?
More than anyone has written down. In practice: instructions to proceed, answers to technical queries, approvals of samples and setting out, changes of sequence, decisions taken quickly when something unexpected is found, prices agreed, dates accepted.
The uncomfortable part is not that this happens. It is that the chat is usually the most contemporaneous and most honest account of the job that exists, and it sits on personal devices belonging to people who may have moved on by the time it matters.
Can a message really form a binding agreement?
Yes, and the clearest authority is recent. In Jaevee Homes Ltd v Fincham (t/a Fincham Demolition) [2025] EWHC 942 (TCC) the Technology and Construction Court held that a demolition contract was formed on 17 May 2023, when a contractor asked whether the job was his so he could start getting organised and the reply was one word: yes.
A formal sub-contract was emailed nine days later, at a different figure and with a different payment regime. The other side never acknowledged it, so it did not apply. Around £195,000 was invoiced, £80,000 was paid, and the difference was litigated on the basis of a contract that existed only in a phone.
None of this is a quirk of new technology. English law has never required most commercial contracts to take a particular form. Offer, acceptance, consideration and an intention to create legal relations can be found in a thread as readily as in a letter. What changed is where the conversation happens.
Two related points are worth carrying:
Brevity is not protection. In Canada, in South West Terminal Ltd v Achter Land & Cattle Ltd, 2023 SKKB 116, a thumbs up emoji sent in reply to a contract was held to be acceptance, and the sender was ordered to pay around 82,000 Canadian dollars. The court asked what a reasonable person would understand the reply to mean, which is the question an English court asks too. A reaction is a reply.
Deleting is worse than keeping. Messages are disclosable, including on personal devices used for work. In Ocado Group plc v McKeeve a solicitor was found in criminal contempt and fined £25,000 for telling a contact to delete a messaging application after a search order had been granted. Whatever a thread contains, destroying it turns a commercial problem into a conduct problem.
Which other industries have already paid for this?
Construction is late to this rather than unlucky. The same pattern has arrived in most sectors where work moves faster than the system built to record it.
| Sector | How it surfaces | What it costs |
|---|---|---|
| Financial services | Business conducted on personal messaging outside monitored channels | Penalties past three billion dollars across more than a hundred firms since 2021, with further settlements in January 2025 |
| Automotive and manufacturing | Engineering changes, deviations and concessions agreed in chat and never entered in the change record | Surfaces in recall and safety investigations, where the question is who approved the departure and when |
| Journalism | Source contact and commitments made on encrypted messaging | Undertakings to sources, and disclosure in defamation claims, from an app designed to leave little behind |
| Healthcare | Clinical decisions and handovers passed between staff on personal phones | Information governance breaches, and gaps in the clinical record when care is reviewed |
| Legal and professional services | Client instructions given informally between formal advice | Scope and retainer disputes, and questions about what was actually advised |
The financial services example is the instructive one, because those firms were mostly not accused of any underlying wrongdoing. They were penalised because the conversation happened somewhere the record could not reach. That is a compliance failure with no bad actor in it, and it is the same shape as the problem on a construction project.
Does banning it work?
No, and it usually survives about a fortnight.
A ban is unenforceable against the people who are not yours, and they send most of the messages. The approved alternative is slower at exactly the moment speed matters, so it loses. Most importantly, a ban does not remove the messages. It moves them into individual chats on personal numbers, which means that when something is tested eighteen months later you are asking former employees to search their own phones for material that costs them nothing to lose and costs you a great deal.
Policy that fights the channel people have chosen is policy that produces less visibility, not more.
What does workable governance look like?
Five things, none of which is an instruction to stop.
- Scope it. Decide which groups are business groups and name them. A group for the job is a record. A group for the five a side is not.
- Say so once, in writing. Tell people that messages in those groups are business records, may be retained and may be produced. It costs nothing and it changes how people write.
- Name where decisions live. A message can start a decision. It should not be the only place the decision exists. The thread is the trigger, not the filing system.
- Capture automatically. This is the one people skip, and it decides whether the rest survives a busy Thursday. If moving a message into the record is a second job done by hand, it gets done for two weeks and then it does not. What good capture looks like in practice, and the chain of custody problem underneath it, is set out in treating your WhatsApp groups as project records.
- Retain properly. Messages and their media, with sender, timestamp and the original file, kept for as long as the matter could run. A screenshot pasted into a document months later is not the same thing, and the other side will say so.
Where does AI Metric come in?
The liability above is not an argument for a better policy. It is an argument for a record, because every one of those cases turned on what could be produced afterwards and by whom.
That is the problem we built Construction Metric around, and the approach generalises past construction: the message, the media at full resolution, the sender and the time, taken into a dated record as it happens, with the original kept alongside. The design constraint is that nobody is asked to change what they do, which is the same reason the processes worth automating first are the ones nobody enjoys doing, and why we treat evidence as the point of the exercise rather than a by-product of it.
If the honest position is that the real record of your business is on people's phones, that is worth addressing before it is tested rather than after. We work through it against your own contracts, your own obligations and the way your people actually communicate, which is most of what our consultancy work is.
Get in touch and we will look at it properly.