It is hard to have a sensible conversation about AI at the moment. On one side are the people who have moved everything onto it and spend their days looking for the next thing to move. On the other are the people who have had enough: enough generated articles, enough chatbots where a phone number used to be, enough “AI-powered” stickers on products that did the same thing last year. That second group has started rejecting anything with AI anywhere near it and you can see why.
We would like to make a case for the middle, because that is where we have ended up.
AI is useful for some things. It reads documents faster than people do, it does not get bored on the four-hundredth certificate and it is good at a first draft. It is also responsible for an astonishing amount of noise. Content for the sake of content. Features for the sake of a press release.
The National Financial Ombud gave a good example of this in March 2026. Complaints that used to be a paragraph or two now arrive as 150-page documents, drafted by AI, quoting legislation and case law that is often wrong and sometimes does not exist. The banks and insurers on the other side reply with AI of their own. In the middle sits an investigator who has to read all of it, and the head ombud, asked how often this had improved a complainant’s chances, could not think of a case. The words got longer. The work got harder. Nobody’s problem got solved any faster.
That is the thing to watch for. Generated content inflates. It makes more work for everyone downstream who has to read it, check it and act on it, and most of that work is meaningless. A tool that produces more of something nobody asked for has not helped, however clever the technology underneath.
So the question we ask ourselves is not “where else can we use AI?” It is two questions: Where would a person be glad to hand this over? And does it do the job without adding fluff or adding to the workload of everyone who comes after? Anything that passes the first and fails the second is noise with a good vocabulary. The list that passes both is short and it is the right list.
“Why buy an HR system when I can just use AI?”
Someone in an IT department is asking this right now and it is a fair question. The tools really have changed. A capable developer with a good AI assistant can produce a working leave system in a week or so and it will look just fine.
We expect this to bring a wave of new HR systems from new companies, from software development houses and even from in-house teams. Some will be good. Software being easier to write is not a bad thing and we are not going to pretend it is.
But an HR system is not mostly software. It is mostly what the software has already been through, and what the people who went through it passed on. Twenty-five years of implementations shape a system in ways that are hard to see from the outside. It is knowing what works and why. It is also knowing which ideas are wonderful in theory and never see the light of day in a real HR department, and why that is. Every one of those lessons was learned alongside a client somewhere and every one of them is now in the product. None of that exists on day one or even year one, however good the tools.
The software is the part a client can see. What they are really paying for is the solution around it: the knowledge, the experience and the ongoing service. Software that changes when legislation changes or that adapts when the business does. Software that addresses needs they did not know they had or never thought software could address. The code is the smallest part of it.
None of this counts against new players. It just means any vendor, including us, should be able to answer a plain question: what have you implemented, for whom, and what did you learn from it?
What we are doing with AI
Version 7 of HR Manage shipped at the beginning of October with AI built in from the start, and built in narrowly on purpose.
We put AI where it saves time that is being wasted on admin. Renewed compliance documents for a few hundred employees go in as batches and AI identifies each document, matches it to the right person and pulls out the expiry date and whatever other fields the requirement needs. A custom report, a certificate or a BI dashboard can be designed by AI once and then used for years without AI touching it again. A paper form that should become part of a workflow becomes a form activity on the system without anyone building it by hand. These are a few of the places where we take dummy work off people’s desks by letting AI do it.
In every one of those the pattern is the same: AI proposes, a person decides and nothing that touches the database without someone looking at it. It is optional and priced separately, so a client who wants none of it loses nothing. It is metered, so you can see what it is doing and what it costs.
There is more coming, at the same pace and under the same rules. We will add things where a person, having used them once, would not want to go back.
That is the middle. It is a less exciting place to stand than either end, but it is where the useful work gets done.
