An AI consultant for an architecture practice finds the admin jobs worth automating, writes rules for client data and drawings, builds one or two workflows, and hands them over. For a ten-person practice, seven consultant days at planning rates of $750 to $1,250 a day come to about $5,000 to $9,000; a single advice session costs a few hundred dollars.
The day rate is the least informative number in a quote. What moves the total is how many days the work needs and how many hours your own people spend in interviews, reviews and testing. Seven days at $750 is $5,250; the same seven at $1,250 is $8,750; and the practice's own 37 hours, valued at $55 an hour, add about $2,000 to either.
Seven consultant days, priced deliverable by deliverable
A first engagement for a practice of about ten people tends to split into five deliverables. The totals below use two planning rates, $750 and $1,250 a day, which the next section builds from published rate guides. Put the rate from any quote you receive into the same table and the comparison still holds.
| Deliverable | What you hold at the end | Days | At $750 a day | At $1,250 a day |
|---|---|---|---|---|
| Task audit and ranked shortlist | Eight to ten admin jobs ranked by weekly hours, data risk and a build-now, later or never verdict | 1 | $750 | $1,250 |
| Policy and tool decision | A two-page AI use policy and a one-page tool note with seat costs | 1 | $750 | $1,250 |
| One workflow, built and tested | A working workflow, its prompts and instructions, and a test log from five real examples | 3 | $2,250 | $3,750 |
| Team session and playbook | A 90-minute session on the practice's own documents and a one-page playbook | 1 | $750 | $1,250 |
| 30-day check and fixes | Review notes, repaired prompts and a decision on the next job | 1 | $750 | $1,250 |
| Total | 7 | $5,250 | $8,750 |
The audit. The consultant spends an hour each with a director, a project architect and the practice manager, then works from one week of the team's own task log: a plain table of what people did, how long it took and how often it recurs. The output is the ranked list. In a practice this size the usual candidates are meeting minutes and action logs, consultant chasers, drawing issue registers and transmittals, first drafts of fee proposals built from past ones, project write-ups for the website, and the question log that builds up during a bid. A good audit also names the jobs to leave alone, such as any note on a drawing that carries professional responsibility.
The policy and tool decision. Two pages, not twenty. The policy says which kinds of material may go into which tools (client names and addresses, signed appointments, consultants' reports, drawings and models), who reads the output before it leaves the office, and when a client or a contract means someone has to be told. The decision note picks the tool for the first workflow and prices the seats. The architecture AI policy checklist goes clause by clause, which a first engagement has no time to do.
The build. Three days looks generous until you count what a first build contains: connecting the steps, a round of failures on real material, a fix, a second round, and writing the instructions so someone else can run it. A common first choice is meeting notes to minutes to action log to consultant chasers, because the input is text and the people who were in the room can check the output. The same-day meeting minutes tutorial walks through that job on its own.
The team session and playbook. Ninety minutes works when it runs on the practice's own documents: last month's minutes, a real chaser, an old fee proposal. The playbook is one page: the approved tools, the three rules from the policy and four prompt cards. Longer sessions with exercises for each role are a separate purchase, priced in what a practice workshop covers and costs.
The 30-day check. The consultant returns after a month, reads how the workflow was used, repairs prompts that drifted, deletes steps nobody follows and recommends either the next job or stopping. It is the easiest day to cut from a quote, and workflows tend to drift when nobody looks at them again.
Where the day rates come from, and what each source leaves out
Only one architecture-specific price list turned up in the sources checked for this tutorial in October 2026, and it is a consultancy's own marketing page. The other pages aimed at practices describe services and publish no fees. The remaining figures come from general AI consulting and freelance rate guides, and each one measures something different.
| Source | What it measures | Published figures | What it leaves out |
|---|---|---|---|
| goLance rate guide, updated October 2026 | Hourly bands for AI consultants by experience | Junior (0 to 2 years) $40-$65; mid-level (2 to 5 years) $65-$120; senior (5 to 10 years) $120-$200; expert (10+ years, niche) $200-$350 | The same page says most AI consultants on the platform charge $12 to $35 an hour, so these bands sit above what most of its own members ask |
| Refonte Learning pricing guide, July 2026 | Day rates for an AI project lead or strategist | Junior $600-$900; mid-level $1,000-$1,400; senior $1,600+ | Written for consultants setting their own rates; it calls the figures approximate and assumes mid-to-large clients |
| Contra, Zapier freelancer page | Hourly rates and fixed prices for automation builders | $30-$150 an hour; $500-$1,500 for a simple two or three app automation; $2,000-$10,000+ for complex workflows; retainers $1,000-$5,000 a month | Prices a build, not advice: no audit, no policy, no handover session |
| Clutch AI pricing guide, 2026 | AI development firms reviewed on Clutch | Most charge about $25-$49 an hour; reviewed projects mostly fall at $10,000-$49,999 | No consulting-specific rates; it gives about ten months as a typical project timeline |
| Main & Machine pricing page, October 2026 | One firm's own published offers for small businesses | Readiness audit $3,500-$8,500 over two to four weeks; implementation sprints $18,000-$60,000; optional support from $1,500 a month | One firm's price list, not a survey; no day rates |
| An AI consultancy's playbook page for architecture practices, October 2026 | One firm's offer to architects | Audit $1,500 flat, called a founding rate with 'standard pricing will be higher'; setup $3,500-$6,500 depending on firm size; an ongoing 'operator' service $800-$1,500 a month | A marketing page that calls itself an example playbook, not a client record, yet carries a live offer; setup depends on firm size, live projects and drawing complexity, with no sizes given |
The planning rates fall out of the first two rows. goLance's mid-level band at eight hours a day is $520 to $960, and its senior band is $960 to $1,600. Refonte's mid-level strategist band, $1,000 to $1,400, overlaps the top of the first and the bottom of the second. The middle of each goLance band rounds to $750 and $1,250 a day, so those are the two rates used throughout. Treat them as planning assumptions for reading quotes, not as a market average: none of the sources surveys what architecture practices actually pay, and each named firm in the table is one price list. A quote under $500 a day sits in goLance's junior band; one above $1,600 is in its expert band, and it should come with a reason why your project needs that.
Two checks against the published fees in the table. Seven days at $750 is $5,250, inside Main & Machine's $3,500 to $8,500 audit range even though the seven days here include a build and a team session as well as the audit; at $1,250 the total is $8,750, just above the top of that range. The same firm prices a build as a sprint of $18,000 to $60,000, well above three planning days, so ask what the build line in any quote includes: acceptance tests, written instructions and how many workflows. The architecture page's audit plus setup comes to $5,000 to $8,000, close to the same band, though it doesn't say what size of practice sits behind its setup range. Contra's fixed prices are for freelancers building automations; the build line here also includes testing on real documents and written instructions, which is why three days at these rates ($2,250 to $3,750) sit above its simple band and at the low end of its complex one.
First-year cost for a ten-person practice, and the payback sum
The consultant's invoice is the only figure anyone quotes, and it is not the whole bill. Three items make up the first year, shown for a practice that pays for four business-plan assistant seats and gives the rest of the team the free tools in its existing office suite.
| Item | Low case | High case |
|---|---|---|
| Consultant, seven days | $5,250 | $8,750 |
| Four assistant seats at $21 a month for twelve months | $1,008 | $1,008 |
| Practice staff time, 37 hours at a blended $55 | $2,035 | $2,035 |
| First-year total | $8,293 | $11,793 |
The 37 hours break down as seven in the audit interviews, three reviewing the policy, ten testing the build, fifteen in the team session (ten people for ninety minutes) and two at the 30-day check. The $55 is an assumed blended internal cost per hour; use your own figure. On seats, Microsoft lists Microsoft 365 Copilot Business at $21 a user a month on an annual term ($25.20 billed monthly) and shows a promotional $18 for purchases between 1 July and 31 December 2026, first year only. The sums use $21 so they stay true after the promotion ends. Ten seats instead of four would add $1,512 a year.
The payback sum is an assumption and deserves to be treated as one. Suppose the practice holds eight client, consultant and site meetings a week, each write-up used to take about 40 minutes, and the workflow cuts that to roughly 15 minutes of checking. That is 25 minutes saved eight times a week, about 3.3 hours; over 48 working weeks it comes to 160 hours, and at $55 an hour, $8,800. Against the $7,285 to $10,785 spent on the consultant and staff time, and net of $84 a month for the seats, the outlay comes back in about eleven to seventeen months. The saving is only real if those 160 hours go to chargeable work or to work the practice would otherwise pay for. Time saved and then left idle returns nothing, which is why the audit's ranking matters more than the day rate.
Advice session, scoped engagement or readiness programme
Consultants sell the same knowledge in three shapes, and the cheapest is sometimes the right one.
| Shape | What you buy | Price | Suits | The catch |
|---|---|---|---|---|
| Advice session | Two to three hours with a consultant | $130 to $600, from goLance's mid-level to senior hourly bands ($65 to $200) | A practice with someone technical enough to build the workflow themselves | Nothing in writing unless you ask for the ranked list and a policy outline |
| Scoped first engagement | Five to eight consultant days: audit, policy, one build, session, check | $3,750 to $10,000 (five days at $750, eight at $1,250) | A practice of five to twenty people with nobody free to build | A quote with no days listed cannot be compared |
| Readiness programme | An assessment across every role, a roadmap and several workflows | $15,000 to $30,000 for the assessment alone, on one construction advisory firm's page aimed at firms of 20 to 200 staff | Multi-office practices with several departments | Scope designed for a larger organisation |
The first row is the right buy more often than a consultant will tell you. A practice manager who is comfortable with spreadsheets and automation tools can build a minutes workflow in a few evenings if someone else has chosen the job, written the rules and named the tool. Three hours at the top of goLance's senior band costs $600 and produces the ranked list; the build then costs the practice manager's time instead of three consultant days. The trade is speed and testing, because nobody outside the practice will check the finished workflow against your meetings.
The third row is the one to question. A construction-focused advisory page from Phos AI Labs (August 2026) lists readiness assessments at $15,000 to $30,000 for firms of 20 to 200 employees, over three to six weeks; that is one firm's price, not a market rate. A ten-person practice receiving a quote in that band is being sold a programme built for a larger organisation. Ask for a line-by-line breakdown and a version that keeps only the audit, the policy and one build. How the day rates and fee models compare across sectors is covered in how much an AI consultant costs a small business.
Three imaginary quotes, compared like for like
A practice sends the same brief to three consultants and gets three differently shaped answers. The method is to find what each quote leaves out of the brief and price the gap, at the quote's own day rate when it gives one and at the planning rates when it doesn't. All three quotes here are invented for illustration.
| Quote | Headline | Left out of the brief | Gap priced | Comparable total |
|---|---|---|---|---|
| A: 'AI strategy and implementation' | $4,500 fixed, no days stated, tools extra | Policy, team session, 30-day check | $2,250 to $3,750 | $6,750 to $8,250 |
| B: day-rate quote | $900 a day, six days, capped at $5,400 | 30-day check | $900 | $6,300 |
| C: 'AI readiness programme' | $14,000 for four weeks | The built workflow | $2,250 to $3,750 | $16,250 to $17,750, including a roadmap and integration planning nobody asked for |
Quote B is the cheapest and the most complete, but the person matters as much as the arithmetic. A $900 day from someone who has run a practice's meeting notes through a test is a different product from a $900 day from someone who hasn't. Ask what they have built and tested for a similar office. Quote A isn't wrong, only unfinished: ask for the days, the deliverables and the cap, and it becomes comparable. Quote C could be good value for a practice of forty. For ten people, ask which parts you can remove.
Here is the same discipline applied to a single line. A vague line reads: 'AI strategy and implementation support: $6,000.' The version worth signing reads: 'Day 1, task audit and ranked list of up to ten jobs with weekly hours and a data-risk note. Day 2, two-page AI use policy and tool note. Days 3 to 5, build and test the meeting-minutes workflow on five of your past meetings; pass mark is 90% of real actions found. Day 6, team session and one-page playbook. Day 7, 30-day review and fixes. $750 a day, seven days, capped at $5,250. Software bought by the practice.' Same money, but the second can be checked.
One page to send before any consultant quotes
Quotes only become comparable when everyone answers the same brief. This is a filled-in version for the ten-person practice used above. Edit the facts, keep the headings.
PRACTICE
Ten people: 3 directors, 5 architects and technologists, 1 practice manager,
1 part-time administrator. House extensions and small commercial fit-outs.
GOAL OF THE FIRST ENGAGEMENT
Return at least 3 hours a week of non-chargeable time without sending client
documents to any tool the practice has not approved.
ALREADY IN USE
Office suite with Teams for calls, a shared drive for project files, a
project management tool for tasks.
OFF LIMITS
Drawings, models and signed appointments stay out of consumer AI tools.
Anything that names a client needs a business-plan tool.
PEOPLE AVAILABLE
Practice manager 6 hours in total, one project architect 4 hours,
each director 2 hours.
DELIVERABLES WANTED
1. Ranked list of admin jobs with weekly hours and data risk
2. Two-page AI use policy and a one-page tool note with seat prices
3. One workflow built and tested on five of our past meetings
4. 90-minute team session and a one-page playbook
5. A review 30 days after the session
QUOTE FORMAT
Day rate, days per deliverable, total cap, exclusions, who buys the software,
what we own at the end and where it is stored.The brief does two jobs. It makes quotes comparable, and it shows which consultants read it. One who asks for the task log, or pushes back on a line in 'off limits', has engaged with your practice; one who replies with a standard proposal hasn't. Send it to three and expect the replies to differ most on the build line and the handover.
Eight questions to put to the consultant before you sign
- Which of these days are yours, and which belong to someone else? A good answer names who does the audit, the build and the session. Be wary of 'our team' with no names, because the person who sold the work and the person who builds it can be different.
- What do you need to see, and can you work from redacted samples? A good answer starts with five past meeting transcripts with client names removed and asks for live access only after the policy is agreed. Be wary of a request for administrator access to the shared drive on day one.
- Which tools will you recommend, and do you earn anything from them? Several automation and CRM vendors run referral programmes that pay recurring commission. A good answer lists tools the practice already licenses first and states any referral income in writing. The checks are laid out in checking an AI consultant for tool commissions.
- What do I own at the end, and where does it live? A good answer puts the workflow in the practice's own accounts, with prompts and instructions written down in plain text. Be wary of anything built in the consultant's account and 'shared' with you.
- How long is a day, and what happens when the build overruns? A good answer says eight hours, gives a cap, and requires written agreement before the cap moves. Be wary of open-ended time and materials on a first project.
- What result would make you advise stopping after the audit? A good answer is specific: nothing on the list saves more than two hours a week, or every candidate touches material that cannot go into an approved tool. Be wary of a consultant for whom the answer is always the full programme.
- How will you test the finished workflow? A good answer uses your own past examples with pass marks agreed beforehand. Be wary of 'we'll check it looks right'.
- What will it cost to run each month, and who can change it once you've gone? A good answer names seat prices from the vendors' own pages and a person in the practice who can edit the instructions. Be wary of a monthly retainer proposed before any running cost is mentioned.
One more belongs in the file you keep for the practice's insurer: ask the consultant to state in writing what they will and won't touch, for example drawings, models and anything that carries sign-off. What to tell the insurer itself is covered in what to tell your professional indemnity insurer about AI.
What the consultant's invoice leaves out
Subscriptions start on the day the workflow does. These are list prices as of October 2026, on annual billing unless stated, from each vendor's own pricing page.
| Running cost | List price | Ten seats for twelve months |
|---|---|---|
| Microsoft 365 Copilot Business | $21 a user a month; $25.20 billed monthly | $2,520 |
| ChatGPT Business, Standard seat | $20 a user a month; $25 billed monthly; two-seat minimum | $2,400 |
| Claude Team, Standard seat | $20 a user a month; $25 billed monthly; two-seat minimum | $2,400 |
| Otter Pro, transcripts | $8.33 a user a month; $16.99 billed monthly | About $1,000 |
| Zapier Professional, 750 tasks | $19.99 a month; $29.99 billed monthly | $240 for one account |
Microsoft 365 business plans already include Copilot Chat at no extra cost, so the free baseline is worth testing before the practice pays for seats.
Support is the other cost that arrives later. Contra's page puts retainers at $1,000 to $5,000 a month, Main & Machine lists optional support from $1,500 a month and the architecture playbook page lists $800 to $1,500. For a practice with one workflow, each is far more than the work needs. A cheaper arrangement is a review day booked a quarter ahead at the agreed day rate, $750 to $1,250 four times a year.
The third cost is rebuilding. A first engagement that built the minutes assistant as a custom GPT faces OpenAI's scheduled shutdown of custom GPTs on 11 December 2026, and the practice then has to move the instructions to a Project or a plugin. Whole products can disappear too: the AI calendar tool Clockwise shut down on 27 March 2026 and deleted its users' data. Prompts and instructions kept as plain text, in the practice's own files, are the cheapest protection against both.
Proving the first workflow worked, with numbers
Agree the pass marks before the build, not after it. For the minutes workflow, a workable set is: at least 90% of the real actions appear in the action log, nothing invented survives the reviewer's read-through, and the average fix takes under ten minutes. Run the workflow on five past meetings whose real action lists someone has already written down. These results are illustrative.
| Past meeting | Real actions | Found by the workflow | Missed | Invented | Minutes to fix |
|---|---|---|---|---|---|
| Client briefing, 52 minutes | 9 | 9 | 0 | 0 | 6 |
| Consultant coordination, 70 minutes | 14 | 12 | 2 | 1 | 12 |
| Site visit, 45 minutes | 7 | 7 | 0 | 0 | 5 |
| Internal design review, 38 minutes | 6 | 5 | 1 | 0 | 8 |
| Approvals call, 30 minutes | 4 | 4 | 0 | 1 | 4 |
| Total | 40 | 37 | 3 | 2 | 35 |
Thirty-seven of 40 is 92.5%, so the first pass mark holds. Both invented actions were caught in review, which satisfies the second mark only because someone reads every log; keep that read-through in the playbook. The average fix of seven minutes (35 divided by five) is under the ten-minute mark and well under the 40 minutes the write-ups used to take.
The coordination meeting is the weak spot: two missed actions and one invented in 70 minutes with several voices. A consultant who sees that result should adjust the workflow, for example by having the chair read the actions aloud at the end so the transcript has a clean summary to anchor on. One who hides it will pass the build and leave the practice to find the problem. The 30-day check exists to re-run this table on new meetings and see whether the numbers held.
Ways to shrink the bill and keep the result
A lean version of the same engagement runs to 4.5 days: audit half a day, policy one day, build two days, session half a day, check half a day. That costs $3,375 at $750 a day and $5,625 at $1,250. The savings come from the practice doing some of the work, not from skipping steps.
- Bring the task log. A week of notes on what people did and how long it took cuts the audit from a day to half a day, saving $375 to $625.
- Supply redacted samples. Meeting transcripts with names removed save the consultant a day of finding and cleaning material, and keep client data out of the build.
- Record the session. Ninety minutes on video, with the playbook, covers the people who join the practice later and removes the need for a second session.
- Start from the tools you have. The tool note can begin with the free Copilot Chat in a business Microsoft 365 plan and only recommend paid seats where the test shows they are needed.
- Buy a cap, not an estimate. A maximum fee with a written change process moves the risk of a long build onto the person who controls it.
Two days stay in whatever the size: the audit, which stops you building the wrong thing, and the 30-day check, which stops the right thing from decaying. If you'd rather work through the ranked list with someone, that is what my AI implementation consultation covers: finding the jobs where AI pays off, automating them or adding AI to the tools the team already uses, and handing over something the practice can run. The contact page has the form.
Questions practices ask before booking an AI consultant
Can a small practice get started for under $1,000?
Yes, by buying one advice session instead of an engagement. Two to three hours at goLance's published mid-level to senior hourly bands of $65 to $200 comes to $130 to $600. Ask for the ranked list of jobs and a one-page policy outline in writing, because the build then falls to someone in the practice and nobody else will test it against your meetings.
Should the consultant's fee include software licences?
No. Buy every subscription in the practice's own name, on its own card, so the accounts, the data and the instructions stay with you if the engagement ends. The quote should show seat prices separately from days, and the decision note should state the monthly bill. A consultant who resells licences may add a margin and may hold the admin login.
How long does a seven-day engagement take on the calendar?
About six weeks in the worked example, because the days are not consecutive: the practice has to supply samples, review the policy and test the build between sessions, and the 30-day check lands a month after the team session. A plan with seven days back to back leaves no gaps for your people to test anything.
Is a free AI audit worth taking?
It can be, if you know where the free part ends. Ask what you will receive in writing, how long it takes, whether it names tools the consultant resells and what the next paid step costs. A free session that ends with a ranked list is useful; one that ends with a proposal for a larger programme is a sales call, which is fine when you know that going in.
Does the consultant need access to our live project files?
Rarely in a first engagement. The audit works from interviews and a task log, and the build can be tested on meeting transcripts or past documents with names removed. Live access should come last, after the policy is agreed, limited to the accounts the workflow uses and switched off when the engagement ends.
Further reads
- Who Owns the AI Workflows a Consultant Builds for You? — What you should own when the consultant leaves, and where it is stored.
- How Much Does AI Cost a Small Architecture Practice per Month? — The monthly subscription view once the first workflow is running.
- Fixed-Scope vs Hourly AI Consulting: Which Protects Your Budget? — Fixed scope or hourly billing, and when each one suits a first project.
- How to Choose an AI Consultant: 20 Questions to Ask First — Twenty questions that apply to any sector, not only architecture.
- What a Typical AI Consulting Engagement Looks Like, Week by Week — What each week of a consulting engagement looks like from the client side.
- AI Automation for Architects: Enquiry to Fee Invoice, Step by Step — A second workflow to consider once the minutes job is working.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: goLance AI consultant hourly rate guide (updated October 2026); Refonte Learning, AI consulting pricing guide (July 2026); Contra, Zapier freelancer hiring page; Clutch, AI pricing guide 2026; Main & Machine pricing page (October 2026); an AI consultancy's example playbook page for architecture practices (October 2026); Phos AI Labs, AI advisory cost for construction firms (August 2026); Microsoft's Copilot Business pricing page; ChatGPT Business, Claude Team, Otter and Zapier pricing pages as checked in September and October 2026.