Splitting the drafting from the judgement calls

Most of a Planning Statement is description, and description is the part that eats architects' unbilled evenings. The useful question when weighing an AI draft against consultant judgement is not whether the machine writes well. It is which sections of the document carry professional liability, because those are the sections worth paying a chartered planner to write, and the rest can be drafted and then checked.

The split is cleaner than it looks. Roughly two thirds of a typical statement records facts that exist somewhere already. The remaining third is argument that depends on a planner's reading of policy, a view on how the council will respond, and a willingness to put a name to a conclusion.

Which parts of a document are description

These sections are reconstructions of information held elsewhere. They need accuracy and completeness, not opinion.

• The site description, surroundings and designations, which come from the title plan, a site visit and the constraint datasets.
• Planning history, taken from the council's register with application references and decision dates.
• The description of development, which has to match the application form and the drawings exactly.
• The development plan policy schedule, listing the adopted plan, its date, the relevant development management policies and any supplementary planning documents.
• The relevant parts of national policy and guidance, cited by edition.
• Consultation undertaken to date and what came back.
• Compliance with the national and local validation requirements.

Getting these wrong causes validation failures and officer irritation rather than refusals, which is exactly why they are the wrong place to spend senior time. They are also the sections where AI drafts get the structure right and the evidence wrong, so the review burden is checking facts against sources rather than rewriting prose.

Which parts are professional judgement

Everything that involves weighing. A planner is being paid for the weighing.

The planning balance is the obvious one. Whether a conflict with a single policy is outweighed by the contribution to housing supply, whether harm to a heritage asset is less than substantial and where it sits within that range, whether a Green Belt case reaches very special circumstances, how much weight an emerging local plan carries given its stage and the extent of unresolved objections. Section 38(6) of the Planning and Compulsory Purchase Act 2004 requires the determination to be made in accordance with the development plan unless material considerations indicate otherwise, and identifying which considerations are material and how heavily they weigh is not a drafting task.

Policy interpretation sits here too. Local plan wording is often ambiguous, and whether a particular policy applies to your site at all can be the whole argument. So can the question of which of two conflicting policies takes priority, or whether a policy is out of date.

Then there is the honest assessment of harm. A statement that concedes a point the officer would have found anyway reads as credible and buys goodwill on the points you are contesting. Deciding what to concede is judgement, and it is the decision clients least want to hear and most need.

Site-specific design reasoning behaves the same way, which is why site judgement cannot be drafted into a design statement from constraint data alone.

Negotiation positions and tactical choices

None of this belongs in the document, and all of it shapes the document.

Whether to raise a weakness pre-emptively or wait to see if it is spotted. Whether to accept an extension of time or let the determination period expire and appeal against non-determination. Whether to submit a section 73 application to vary a condition or a fresh application that reopens everything. Whether to offer a condition the officer has not asked for, to close down a refusal reason. Whether to ask for the application to go to committee, or hope it stays delegated.

Those choices change what the statement says. A statement written to support a committee presentation reads differently from one written to be read by a single officer under time pressure, and both differ from one written with an appeal in mind. A drafting engine cannot make that call because it does not know the local political context or what the client can tolerate.

Risk assessment on contested policy

The piece of work most often skipped is a short internal note, never submitted, setting out where the application is vulnerable.

A usable version covers, for each contested point: what the policy says, what the council's recent interpretation appears to be, what evidence supports your reading, what a refusal reason would say, and whether the point would succeed on appeal. Rate each one. Then decide which to argue in the statement, which to address with a condition or a design change, and which to leave alone.

This is also where an appeal strategy starts. Where a point is likely to fail at officer level but succeed at appeal, the statement needs to lay the groundwork without signalling that an appeal is planned. Doing that well under time pressure is hard, and there is a separate discipline to drafting to an appeal deadline without cutting corners.

How to brief a consultant on a draft

Handing a planner a finished draft and asking them to review it is the wrong instruction. It invites a line edit when you want a judgement.

A better brief:

1. Send the draft with the descriptive sections complete and sourced, and say plainly that they are drafted material to be checked, not authored.
2. Send the list of items flagged as unconfirmed, so the planner can see what is missing rather than discovering it.
3. Ask specific questions. Does the heritage balance read correctly. Is the housing land supply position still as stated. Which of the three policy conflicts should we concede.
4. Ask for the planning balance and the conclusion to be written by them, in their words, over their name.
5. Agree who signs the document and what their sign-off covers.

That last point matters more than fee. A reviewer who signs a document is accepting responsibility for the whole of it, so they will reasonably want to re-verify the description, which eats the time saving unless the sourcing is already visible. Record the source against each fact and the review gets faster and less expensive.

Fee structures that reflect the split

Hourly rates applied to drafting reward slow writing. Where the description is already drafted, the honest structures are:

Practices running the Professional tier tend to split it this way: the practice produces and owns the descriptive document, and the planner is engaged for the judgement, the negotiation and the sign-off. The reviewers listed in the professional directory can take a draft and provide that review and sign-off, which keeps the liability where it belongs.

What the client should receive

Three things, and clients rarely get all three.

The document, with the policy schedule showing the edition and date of every source cited, so it can be checked in two years when someone asks whether the plan had changed. A short note naming who authored which sections and what the sign-off covers. And the risk note, in plain language, saying where the application might fail and what happens then.

Version discipline is the quiet failure here. Where sections are regenerated or revised, a client can end up advised on one version and submitting another, which is why version control when report sections are regenerated deserves a process rather than good intentions.

PlanScope produces AI-assisted drafts that are structured to UK planning policy, cite the policies they rely on, and insert flags where site facts are missing. Reports should be reviewed before submission, and PlanScope does not guarantee planning approval. For specialist technical reports, qualified professional input may be required. This is general guidance, not legal advice; check the position with your local planning authority or a qualified professional before relying on it.