AI Opportunity Audit
3 September 2026
Halloway Studio
The recommendation
Start with proposal first drafts. It is the largest single block of senior time and the material needed to do it well already exists.
Hours back each month
31 hrs
Net value per month
$1,487
Cost of a year of delay
$17,844
Start here
This week, pull the last twenty proposals into one folder and time how long the next one takes end to end. That single measurement is what every claim in this report will be judged against.
Executive summary
Halloway Studio is carrying roughly 96 hours a month of work that does not require senior judgement but currently receives it. The largest concentration is proposal production, where a partner drafts from scratch each time despite a decade of comparable proposals sitting in Dropbox. Three interventions are worth making this quarter and four are not, mostly because the underlying material is too disorganised to build on or because the volume does not justify the setup. Acting on the recommended three returns about 31 hours a month at a net value of $1,487, with setup recovered inside four months. The binding constraint is not budget or tooling. It is that nothing is currently measured, which means a year from now you would have no way to prove any of this worked.
How the business runs today
Twelve people, three of them partners, running client work through Airtable with files in Dropbox and finance in QuickBooks. Work enters through the founder, who still writes most new business material personally. Approvals route back through the same person, so the practical capacity ceiling is one individual's calendar rather than the team's. Nothing is badly run, which is why the inefficiency has survived: every individual step looks reasonable, and the cost only becomes visible when the hours are totalled.
Workflow as declared
Owner
Hrs / mo
Cost / mo
Proposal writing and first drafts
Founder
26
$1,248
Client brief assembly from kickoff notes
Account leads
22
$1,056
Invoice preparation and chasing
Producer
16
$768
Weekly status reporting
Account leads
18
$864
Meeting notes and follow ups
Whole team
14
$672
Total declared load
96
$4,608
Priority matrix
Business value
Readiness and feasibility
Pursue
Pilot first
Not now
Bubble size shows hours recovered per month
Ranked by weighted priority
Brief assembly from kickoff notes
74
Weekly status reporting
51
Weighted priority score, 0 to 100
Opportunities, scored
Opportunity 01
Pursue
Proposal first drafts
What happens today
The founder writes each proposal from a blank document, pulling structure from memory and occasionally locating a comparable past proposal by searching Dropbox filenames. A typical proposal takes between four and seven hours across two sittings, and roughly one in three is for work the firm has proposed before in near identical form.
What we would change
A drafting assistant grounded in the last two years of accepted proposals produces a structured first draft from a short brief: scope, deliverables, timeline and a pricing skeleton in the firm’s own language. The founder edits rather than writes, which is a fundamentally different cognitive task and takes roughly a quarter of the time.
How it gets built
Consolidate accepted proposals into a single named folder, then a retrieval layer over them feeding a drafting prompt with the firm’s standard section order. Runs inside Google Workspace, so no new document home is introduced. Configuration rather than engineering, around two days of setup including a review pass on the outputs.
How we prove it worked
Median hours from brief received to draft ready, measured across the next ten proposals against the baseline captured in week one.
Main risk
Draft quality depends entirely on the archive being genuinely representative. If the last two years include weak proposals, the assistant will reproduce their habits.
Scoring detail
Business impact
9 / 10
Revenue potential
8 / 10
Feasibility
8 / 10
Implementation complexity
4 / 10
Cost and effort
3 / 10
Readiness of this business
7 / 10
Differentiation
6 / 10
Risk and dependencies
3 / 10
Google Workspace
Claude
Custom retrieval layer
Opportunity 02
Pursue
Brief assembly from kickoff notes
What happens today
Kickoff calls are recorded inconsistently and notes live in whichever tool the account lead prefers. Briefs are then rewritten from those notes, and information that was captured verbally often has to be re-requested from the client, which costs credibility as well as time.
What we would change
Call recordings are transcribed automatically and turned into a structured brief against the firm’s existing template, with unanswered fields flagged explicitly rather than silently omitted. The account lead reviews and confirms instead of reconstructing.
How it gets built
Transcription on the existing call platform, then a templating step writing into Airtable where project records already live. The flagging of missing fields matters more than the drafting: it converts an invisible gap into a visible task.
How we prove it worked
Count of client information requests made after kickoff, per project, compared against the trailing three months.
Main risk
Requires the discipline of recording every kickoff call. Two account leads currently do not, and no tool fixes that.
Scoring detail
Business impact
8 / 10
Revenue potential
5 / 10
Feasibility
8 / 10
Implementation complexity
4 / 10
Cost and effort
3 / 10
Readiness of this business
7 / 10
Differentiation
5 / 10
Risk and dependencies
4 / 10
Airtable
Transcription service
Claude
Opportunity 03
Pilot first
Weekly status reporting
What happens today
Each account lead assembles a status update per client from Airtable records, Slack threads and their own recollection. The format varies by person, and clients occasionally receive contradictory information about the same timeline.
What we would change
A drafted status update generated from Airtable state, which the account lead corrects and sends. The value here is consistency across clients more than the hours saved, which is why it is worth piloting on two accounts before extending.
How it gets built
Scheduled read of Airtable project state into a fixed report format. No new data entry is introduced, which is the reason this is feasible at all: it is reporting on records that already exist rather than asking anyone to maintain new ones.
How we prove it worked
Whether all accounts receive an update in the same format on the same day for four consecutive weeks, plus hours spent per lead.
Main risk
Airtable records are only as current as the team keeps them. A confident weekly report built on stale data is worse than no report, so this needs a data freshness check first.
Scoring detail
Business impact
6 / 10
Revenue potential
3 / 10
Feasibility
7 / 10
Implementation complexity
5 / 10
Cost and effort
4 / 10
Readiness of this business
6 / 10
Differentiation
4 / 10
Risk and dependencies
5 / 10
Airtable
Claude
What we advise against for now
These came up in scoring and did not earn a place in the roadmap. They are listed so the decision is visible rather than quietly dropped.
Not now
Client-facing chatbot on your knowledge base
Your reusable material is scattered across individual inboxes and inconsistently named Dropbox folders. A retrieval system built on it would answer clients confidently and sometimes wrongly, which costs more than the time it saves. Worth revisiting once the proposal archive consolidation in days one to thirty is complete.
Not now
Automated invoice reconciliation
At your transaction volume the build cost does not clear. Payback lands beyond eighteen months even on optimistic assumptions, and QuickBooks already handles the mechanical part. The sixteen hours a month here is chasing, which is a collections problem rather than a data problem.
Not now
Second attempt at an AI note-taker rollout
You bought one last year and it lapsed within a month. Nothing about the current tools has changed that outcome, because the binding constraint was adoption rather than capability. Transcription is included inside the brief assembly work above, where it has an owner and a reason to exist.
Not now
Custom internal search across all systems
This is the intervention people ask for most and it is nearly always premature. It requires your Dropbox, Airtable and Notion content to be consistently structured first. Attempting it now would produce a search tool that returns nothing useful and permanently damages the team’s appetite for the next thing you try.
The financial model
Every figure here is derived from one number you supplied, a blended cost of $48 per hour, applied to the hours each opportunity recovers. Tool subscriptions are already deducted. Nothing is estimated in dollars by the model itself.
Conservative
$892
Assumes 60 percent of the hours are actually recovered, which is normal in the first quarter.
Expected
$1,190
Assumes 80 percent recovery once the team has settled into the new process.
Full
$1,487
All recommended opportunities delivered and fully adopted.
31 hours recovered at $48 equals $1,488 gross, less $1 of monthly tooling, giving $1,487 net per month, or $17,844 a year. One off setup across the recommended items is $5,600, recovered in 3.8 months.
Baseline to capture before building
Everything above is an estimate until these are measured. Record today’s figure before any build starts, because a saving that was never baselined cannot be claimed afterwards.
Median hours from brief received to proposal draft ready
Target under 90 minutes
Time the next three proposals end to end with a stopwatch, including the interruptions. Do not reconstruct it from memory afterwards.
Figure today
Client information requests made after kickoff
Target 50 percent reduction
Count them in the trailing three months of email for four representative projects. Takes about forty minutes.
Figure today
Founder hours per week on new business material
Target 8 hours or fewer
One week of honest calendar and time tracking, categorised into writing, editing and calls.
Figure today
Ninety day roadmap
Days 1 to 30
Measure, then consolidate
›Capture the three baseline figures above before touching any tooling.
›Consolidate accepted proposals from the last two years into one named folder with a consistent naming convention.
›Audit which kickoff calls are actually being recorded and agree that all of them will be.
›Confirm the blended hourly cost with your accountant so the model in this report can be trusted or corrected.
Days 31 to 60
Build the drafting layer
›Stand up the proposal drafting assistant against the consolidated archive.
›Run it in parallel with the current process for the first five proposals, comparing output quality rather than replacing immediately.
›Configure transcription and brief templating into the existing Airtable project records.
›Have the founder review five drafts and adjust the prompt against their actual objections, not hypothetical ones.
Days 61 to 90
Measure again, then decide on the pilot
›Re-measure the three baseline figures and compare honestly, including cases where nothing improved.
›Pilot weekly status reporting on two accounts only.
›Run a data freshness check on Airtable before extending status reporting any further.
›Decide explicitly whether to revisit the client-facing chatbot now that the archive is consolidated.
Recommended tooling
Claude
from $20 per user per month
Drafting and templating across proposals, briefs and status reports.
Google Workspace
already in use
Remains the document home, so no migration is required.
Airtable
already in use
Stays the source of truth for project state and feeds the reporting layer.
Transcription service
roughly $15 to $30 per month
Converts kickoff calls into structured input for brief assembly.
Risks and how they get handled
HIGH
Archive quality determines draft quality
The drafting assistant is only as good as the proposals it learns from, and nobody has assessed whether the last two years are representative.
Mitigation Have the founder mark fifteen proposals as strong or weak during consolidation, and ground the assistant only in the strong set.
MEDIUM
Adoption risk on a team that has abandoned a tool before
Your own account of the note-taker suggests new tools lapse when nobody owns them.
Mitigation Each intervention above has a single named owner and one measurable figure. Nothing is rolled out to the whole team at once.
MEDIUM
Client confidentiality on proposal content
You hold signed NDAs with three accounts, and proposal material may reference their commercial terms.
Mitigation Exclude NDA accounts from the training archive and confirm data retention terms in writing before the day thirty one build.
LOW
Tool pricing drift
Per seat pricing on assistant tools has moved repeatedly over the past two years.
Mitigation The recommended stack sits well inside your stated monthly appetite, leaving room for increases without changing the payback conclusion.
What happens next
Nothing in this roadmap requires outside help. The first thirty days are measurement and file consolidation, which your producer can run. The build in days thirty one to sixty is configuration rather than engineering, and the tooling named above is chosen so that it stays inside your existing stack. If you want a second opinion on the scoring before committing, the eight dimensions are printed for every opportunity so any competent operator can check the reasoning against your own knowledge of the business.
Method and assumptions
All hours in this audit are client estimates supplied during intake and have not been independently observed, which is why the baseline section exists. The blended hourly cost of $48 was supplied by the client and applied uniformly, though in practice partner time costs more than producer time. Tool prices were verified at the time of writing and are subject to change. Adoption assumptions are drawn from the client’s own description of how the team has behaved with previous tools. Nothing here has been tested against live data, and the proposal archive condition in particular should be verified before the day thirty one build begins.
This is a sample. Yours would be about your business.
Halloway Studio is fictional. The format, the scoring, the arithmetic and the refusals are exactly what you receive.
Start my audit
