What a Planning Consultancy Loses to Manual Document Assembly, Per Project

The fee went somewhere, and it was not into writing

A principal at a planning consultancy knows the feeling. The application went in on time, the client was happy, and the job still came in under. Not badly. A day here, a day and a half there, on a fixed fee that was priced for a straightforward submission. Nobody can point to where the hours went, because nobody was doing anything that looked wasteful. They were writing planning statements, which is the job.

So you go looking for a fix and the first page of results gives you two things. The first is a council's free template for a design and access statement. The second is a tutorial on getting an AI to draft a report in twenty minutes. Both are built on the same assumption: the document is the unit of work, and producing it faster is the saving.

We have looked at enough of these firms to say the assumption is wrong. The document is not where the fee leaks. The fee leaks in the space between documents.

The templates are free, and that is the tell

If document production were the expensive part of a planning submission, councils would not be giving it away.

The Planning Portal's own guidance says a design and access statement should be proportionate to the complexity of the application, and that for most straightforward applications it may only need to be a page long. Cotswold District Council publishes Word templates for three types of statement, meant to be edited directly and submitted. Other authorities do the same. The format is standardised, the headings are known, and a competent planner can fill one in from a site visit and a policy map without much thought.

When something is free and standardised, it has stopped being the constraint. The margin on a planning submission was never in the typing. It sits in the fact that the same site does not get described once. It gets described four or five times, in documents with different purposes, by people who each remember it slightly differently.

One site, five restatements, four memories

Walk through a routine householder or minor application and count the places the same facts appear.

The site address, the site area, the existing use, the proposed floor space, the number of units, the conservation area or listing status, the flood zone, the relevant local plan policies, and the consultation history. All of that goes into the application form. It goes again into the planning statement, in prose. It goes into the design and access statement, framed around design response. It goes into the covering letter, summarised for the case officer. If there is a heritage statement or a flood risk note, it goes there too.

None of that is drafted from a single source of truth. The form was filled in by the technician from the client's brief. The planning statement was written by the planner from the site visit notes. The design and access statement was adapted from the architect's own version. The covering letter was written last, by the principal, from memory of the whole thing. Four people, four recollections of the same site, four documents that were consistent on the day each one was written.

Assembling those documents is cheap. Any firm can do it in a morning. Then somebody changes a number.

The cost is reconciliation, not assembly

The floor area gets revised after a client conversation. The architect issues a new drawing set with one fewer parking space. A policy citation turns out to refer to the superseded local plan. Each of those is a two-minute correction in the document where it was noticed. It is also a silent error in every other document that restated the same fact from somebody else's memory.

Nobody is paid to check that the covering letter still agrees with the form. That work happens anyway, usually at the worst time: the evening before submission, or after validation, when a case officer notices the site area on the form does not match the statement and the whole package bounces.

The bounce is not rare. Planda's reporting on invalid applications cites Tewkesbury Council invalidating 45 percent of the applications it received in a year, and the causes listed across councils are the mundane ones: missing plans, unsigned forms, drawings without labels, omitted statements. Those are not planning failures. They are reconciliation failures, and each one puts the project back through the same set of hands to find which document drifted from which.

The industry's own data says where this lands. The 2026 Architecture and Engineering Industry Benchmark, published by Factor, reports 73 percent of firms naming scope creep as their biggest budget killer, 56 percent reporting overruns on more than a tenth of their projects, and 40 percent not tracking project profitability in real time. Its respondents skew small, roughly 80 percent under twenty staff, which is exactly the size of most planning consultancies. Those firms are fully busy. They cannot see the overrun until the project closes, and when they look for the cause they find nothing that resembles waste, because reconciliation looks like diligence.

We do not have a sourced figure for hours spent reconciling per planning submission, and we are not going to invent one. What we can tell you is the shape of the number. It scales with how many documents restate the same facts and how many revisions arrive after the first draft. A firm that submits five documents and takes three revisions is doing fifteen reconciliation checks per project, and doing most of them by rereading.

Why the AI tutorial makes it worse

This is where the second page-one result needs a warning. A tool that drafts a planning statement in twenty minutes is a tool that produces a fifth restatement of the site, from a sixth memory, faster than before.

Generation lowers the cost of the cheap step. It does nothing to the expensive one, and in practice it raises it, because a document that was produced in twenty minutes gets less scrutiny than one that took a day. The number of things that have to agree goes up. The number of people who know why any given number is in any given paragraph goes down.

We have watched firms adopt drafting tools and conclude, six months later, that the fee did not improve. It did not improve because they automated assembly and left reconciliation exactly where it was. The mechanism is the same one we described in the automation ROI math vendors show you, recalculated. Hours saved on the visible step get counted. Hours created on the invisible one never appear in the model.

What the fix looks like, at altitude

The firms that have stopped losing fee to this did not buy a document generator. They changed what a document is.

In the well-run version, the site facts exist once. The address, the areas, the designations, the policies, the constraint schedule, and the drawing revision each live in a single place that the planner owns. The form, the statements, and the letter are views of that record, not independent retellings of it. When the architect issues a new drawing, the revision changes in one place and the question "which documents still say the old number" has an answer that does not depend on anyone's memory.

That is a structural change, not a software purchase, and it can be done with unglamorous tools. Some of the firms we have seen do it well use a shared schedule and a discipline about which document is the master for each fact. The ones that bought a platform first and hoped the structure would follow are the ones still reconciling by rereading. This is the same lesson a multi-location retailer learns when its systems each keep their own copy of the stock count: the customer notices once, the back office pays every night.

There is a decision underneath this that only the principal can make. Some restatement is genuinely required, because the case officer reads the covering letter and the committee reads the statement and they want different things. Some of it is habit. Deciding which is which, per application type, is the work, and it is worth doing before any tool conversation, because a tool that automates the habit preserves it.

If you want a quick sense of whether your firm has this problem, take the last submission that was invalidated or bounced back with a query. Find the fact that was wrong. Then count how many documents in the package stated that fact, and how many of them agreed. If the answer is "all of them, except the one the case officer read," you have found your fee.

We have said this about other kinds of paperwork, including what a permit actually costs the authority on the other side of the counter. The pattern does not change with the industry. The document that is cheapest to produce is the one whose cost is hardest to see.

Where this stops being an article is where the answers are specific to your practice. Which documents restate which facts, how many revisions a typical application actually takes, whether your invalidations trace back to drift or to genuinely missing material, and what your realised fee per submission looks like once the reconciliation hours are counted: those live in your last twelve projects, and finding them takes a week of sitting with your planners and your job costing.

If your submissions go in on time and your fixed fees still come in under, talk to us. We will find where the same facts are being restated, count what keeping them agreed is costing you per project, and tell you whether the fix is a template, a structure, or nothing to buy at all.