The line a data tool can't cross

A land data tool earns its subscription. It finds the parcel, tells you who owns it, flags the constraints, surfaces the comparable that used to take an afternoon to dig out by hand. Used well, it turns three days of desk research into an hour. None of what follows is an argument against that.

There is a line in every deal where the question changes. Up to that line, the work is "what is this site". Past it, the work is "what do we do about it, given who we are". The second question is where land firms compete, and it is the one a data tool cannot answer, because answering it depends on things the tool does not hold.

A shared stack holds shared facts

A data platform is the same for every firm that pays for it. Ownership, constraints and comparables are facts, equally true for you and the firm in the next town. The tool levels the field on facts, which is useful, and it also means facts stop being an edge the moment everyone has the same login.

If your advantage was knowing where the sites are, a subscription has already eroded it. Everyone can see the map now.

Your edge is in how your firm reads a site

What a competitor cannot buy is how you read what the map shows. Your appetite for risk. Your funding. The deal you did two villages over that taught you what the highways authority will actually accept. The kind of owner you handle well and the kind you leave alone. None of that lives in a database. It lives in your past deals, your correspondence, the reasons you backed one site and walked from another.

That is the real asset, and it is the one most firms do the least to protect. It sits in people's heads and sent folders, available only when the right person happens to remember.

Where the grunt work actually is

Most of the hours past the line are not dramatic judgement calls. They are admin that carries judgement: drafting the offer letter in your house style, chasing the owner on the right cadence, running the appraisal on your assumptions rather than generic ones, keeping the file current as things move. A data tool was never built for this work, and a junior doing it by hand is slow and costs you a salary.

Where Harold comes in

This is the part Harold is built for. It is a property AI trained on your firm's own deals, so it works from how you read a site rather than a generic template, and it lives in the tools you already use. It does the judgement-adjacent grunt: drafting in your voice, chasing on time, appraising on your numbers, keeping the file straight. The decision stays with you, and Harold shows its working so you can check it.

It sits alongside whatever data tool you already pay for, rather than replacing it. One finds the site. The other helps you act on it the way your firm acts.

We are honest that this is early. Harold is shaped on the deals of working UK land operators and configured on your own sites before you commit, so the judgement it reflects is yours, not a vendor's idea of the market.

The question to ask of any tool

Does it know how your firm wins, or only what is on the map. The map is a commodity now. How you read it is the part you still own.

If you want to see where your firm's judgement is locked in people's heads and inboxes, the land newsletter goes deeper, and a short AI audit will show you the first place it is leaking.