Andrew Blair / Practical AI for professional firms

Blog ·

Faster Fee Proposals for Architects, Built on the SACAP Guideline

How AI drafts a fee proposal on the SACAP guideline — faster, still checked by a person. What it can and can't do for your practice.

A new enquiry comes in, you have a good call with the client, and then the momentum stalls at the fee proposal — a document that has to cover scope, exclusions, stages and a number, written from something close to a blank page, usually late at night after the site work is done. It’s not hard work exactly. It’s just slow, repetitive, and it’s the one document that decides whether the job happens at all. I run AB Architecture, a residential practice in Knysna and Cape Town, and fee proposals were one of the first things I rebuilt around AI — because they’re the clearest case of admin that shouldn’t need to start from scratch every time.

What the SACAP guideline actually requires

Before AI comes into it at all, a proposal in South Africa has to sit within the South African Council for the Architectural Profession’s guideline for professional fees. The current guideline, gazetted under the Architectural Profession Act, describes two methods of fee calculation — project cost-based and time-based — and both are meant to be read alongside the framework guideline rather than in isolation. SACAP also publishes a fee calculator built on the current gazetted figures.

That’s the part that doesn’t change no matter how the proposal gets drafted: the guideline sets the basis, and the practice still has to apply it correctly to the actual job — scope, complexity, site conditions, stage breakdown. AI doesn’t get to skip that step or decide the number. What it can do is remove the blank-page problem and the retyping.

How I draft a proposal now

I describe the client and the job out loud — who they are, what the job is, roughly what’s involved. A system I built drafts the proposal in the practice’s house style, using the current SACAP guideline, our standard terms and exclusions, and what we’ve charged on similar jobs before. Before quoting, it can check the property’s zoning and how much has already been built on the site, so the fee reflects the real scope rather than a guess at it. Once the proposal is sent, it watches for a reply and drafts a polite follow-up if the client goes quiet — a small thing, but it’s the difference between a proposal that gets chased and one that gets forgotten.

By my own estimate, that’s brought proposal drafting down to roughly a fifth of the time it used to take — call it 5–6× faster. That’s not a benchmark or an industry figure; it’s what changed in my own week, and it doesn’t include the time I still spend reading the draft against the actual brief before it goes out, which is real and matters more than the drafting speed.

Site conditions are part of getting the scope right too. If a property has servitudes or title deed conditions that restrict what can be built, that affects the real scope of work — and therefore the fee — before a single drawing is started. That’s the kind of check I’d rather catch at proposal stage than discover halfway through a project.

A checklist for a stronger fee proposal

Whether or not AI is involved in drafting it, this is roughly what I check before a proposal goes out:

  1. Confirm the fee basis — project cost-based or time-based — and make sure it’s actually the right fit for how well-defined the scope is.
  2. Check the property, not just the brief. Zoning, existing coverage and any title deed conditions can all change the real scope before you’ve drawn anything.
  3. Match the stage breakdown to the guideline’s structure, not a generic template — this is where AI-drafted proposals go wrong if they’re built from a one-size-fits-all fee template instead of the current gazetted guideline.
  4. State exclusions clearly. Most fee disputes I’ve seen come from what wasn’t excluded, not what was charged.
  5. Read it as the client will. A proposal that’s technically correct but reads like a form letter doesn’t win the job.
  6. Have a person approve it before it’s sent — every time, without exception.

Limits — what AI doesn’t decide

AI drafts faster. It doesn’t decide the fee, and it shouldn’t be left to. The guideline gives a basis, but applying it to a genuinely unusual site, a complicated brief, or a client relationship with its own history is still judgement — the kind that comes from having written enough proposals to know which jobs run over and which don’t. I check every drafted proposal against the guideline and the actual brief before it goes out; nothing leaves the office as a draft-turned-final without that step. If a system like this is ever built for a client firm, that review step is non-negotiable, not optional — it’s built into how I’d set it up, not left to the client to remember.

The other limit worth being honest about: none of this is legal or regulatory advice, and I don’t present it as compliance with anything beyond following the current published guideline as it applies to the fee basis itself. Anything involving client or project data — including what a proposal-drafting tool has access to — is worth thinking through properly with your own advisors before it touches real client information.

Getting started

If proposals are eating more of your week than they should, the fix usually isn’t a generic template — it’s a system that already knows your house style, your standard terms and the current guideline, and that still leaves you checking every word before it goes out. That’s the same shape of system I use for my own fee proposals, and it’s one of the workflows I build for other firms; see the fee proposal case study for the fuller description and Services for how a build like this gets scoped and priced.

If you’d like to talk through what your own proposal process looks like, book a free 30-minute call — no charge, and I’ll tell you honestly whether this would actually save you time.

Tell me what eats your week.

Book a free 30-minute call