A walk-through of a real (anonymized) AI readiness assessment — what we looked at, what we found, and what we recommended.
'AI assessment' has become one of those phrases that means everything and therefore nothing. So instead of defining it abstractly, here is what one actually looked like for a 40-person specialty services company we'll call Northbridge. Details are anonymized; the numbers and findings are real.
The setup
Northbridge came in with a familiar story: 'We feel like we should be doing something with AI, but every vendor pitch sounds the same and we don't know what's actually worth it.' Revenue around $9M, healthy margins, growing 20% a year, and drowning in operational drag.
What we actually looked at
An assessment is not a vendor demo. Over two weeks we did the following:
- Interviewed nine people across sales, operations, finance, and delivery — 45 minutes each.
- Sat with three frontline employees and watched them do their actual jobs.
- Inventoried every tool in use: 23 SaaS subscriptions, 6 spreadsheets-of-record, 2 internal apps.
- Mapped the journey of a typical job from first inquiry to final invoice — 47 distinct steps.
- Quantified time spent on each step using a mix of timesheets and direct observation.
What we found
Three findings ended up driving the recommendation. None of them were 'add a chatbot.'
Finding 1: 31% of operations capacity was copy-paste
Ops coordinators spent roughly one full day a week moving data between the CRM, the scheduling tool, and QuickBooks. Not analyzing it — literally retyping it. At loaded cost, that was about $148K a year of payroll spent on reformatting.
Finding 2: Quote-to-cash had a 9-day average drag
From the moment a job was approved to the moment the invoice went out, the average was 9 calendar days, with no single owner. The delay wasn't laziness — it was the workflow living in three tools that didn't talk to each other.
Finding 3: The estimating workflow was the real bottleneck
Sales was 'fast' but estimating took 2–4 days per quote because the senior estimator had to manually pull historical data from past jobs to price new ones. The data existed — it was just trapped in PDFs.
What we recommended
A 90-day sequence, in this order:
- 1Build a small internal app that owns the job record end-to-end and syncs into QuickBooks, eliminating the copy-paste loop. (~6 weeks)
- 2Layer an AI extraction step over historical estimate PDFs so the estimator can search past jobs by parameters in seconds instead of opening files one by one. (~3 weeks)
- 3Add automated nudges and ownership rules to collapse quote-to-cash from 9 days to a target of 3. (~ongoing)
What it was worth
Conservatively modeled, the three changes recovered about $230K/year in operating capacity, accelerated cash collection by roughly $400K of float, and — the part the CEO actually cared about — gave the senior estimator 60% of his week back to mentor two juniors.
That's what a real assessment produces: a small number of specific, prioritized bets with numbers attached. Not a slide deck about 'the AI opportunity.'
