AI Policy Task Force · Company Survey — Detailed Read
Beneath the headlines.
The company-wide numbers held up — but the interesting story is in the cuts: office vs. field, users vs. non-users, and where AI actually lives today. This is the task-force working view for Meeting 2.
55 responses · closed Sept 21 · for task-force discussion
The one-screen recap
Where the company landed
Feelings led with Curious (31) and Cautious (31), tied; negatives barely registered. The shared worry — confidently-wrong output used without checking — mirrors the task force's own intake almost word for word. The public version of this is on the survey site; what follows is the part worth the meeting.
Full company-facing results: aisurvey.collage-usa.com/results
The biggest divide
Office and field aren't in the same place
The single most useful cut in the data. The field is meaningfully more cautious — and asks for something different.
| Office (n=38) | Field (n=13) | |
|---|---|---|
| Use AI daily | 10 (26%) | 0 (0%) |
| Use it weekly or more | 17 (45%) | 3 (23%) |
| Top feeling | Curious (66%) | Cautious (54%) |
| #1 thing that would help | Training | Clear rules on what's OK |
| Hesitated or unsure if allowed | 16 (42%) | 7 (54%) |
In the office
n = 38 · curiosity leads
In the field
n = 13 · caution leads
What it means for rollout: a single company-wide pitch will miss half the company. The field is rules-first and hands-off today — meet them with clear guardrails and practical, hands-on examples, not enthusiasm. The office is already leaning in and wants training and room to go further. Same policy, two on-ramps. (Field n=13 is small — read as a signal, confirm as more field voices come in.)
The 17 who haven't started
Cautious, not hostile — and winnable
Nearly a third of respondents (17) don't use AI at all yet — and they're not just the field: Operations/Admin (5), Project management (4) and Field (4) lead the group. Their dominant feeling is Cautious (11 of 17), with only a handful skeptical. This isn't resistance; it's people waiting for permission and a runway.
What the non-users say would help them start:
The on-ramp writes itself: a short intro session, role-specific examples, and a named person to ask. This group is the clearest ROI for the training plan — turning 17 cautious non-users into confident, safe users.
The governance picture
The heaviest users are the least contained
The company-wide stat — 31 of 38 users on personal accounts — gets sharper when you look at the daily users: of the 11 people using AI every day, only 3 are on a company account. The other 8 run on personal or mixed accounts. The people getting the most out of AI are doing it furthest from any company control.
The decision this forces: ChatGPT on personal accounts is where the real usage is — including proprietary and client context that several people flagged as their top worry. One respondent noted Copilot keeps queries inside Collage's Microsoft tenant. Which tools do we approve, on which accounts, and how fast can we stand them up? This is the most concrete, most urgent call in front of the group.
In their own words
The themes, with the voices behind them
All anonymized; screened for names, clients and projects. Grouped by the concern or case they make.
Verify before you rely on it — the near-universal worry
"AI at Collage needs to be a tool to cut out the tedious work, not provide the answer to a problem."
"People being too trusting, not reading and editing the deliverables… trust first, then blaming AI when things go wrong."
"They use AI to build a schedule, and they cannot review the result because they don't know how to build one. Is it right or is it wrong?"
Data security & proprietary information
"Giving proprietary info out to the masses. Getting incorrect or incomplete info and making decisions based off of it."
"Direct all Collage AI requests to Microsoft Copilot — while connected to the company portal, those queries stay protected by our Microsoft account. Third-party options put our queries at risk."
The human craft — don't let it dull us
"Our industry requires minds that are good at critical thinking and problem solving. We should constantly be developing our own intelligence."
"Used the right way, it can help grow the company… It doesn't hang drywall, pour concrete, or manage a sub. We still run the show."
Ready to go — the enthusiasts
"I have been able to compile documents in two–three minutes that might otherwise have taken an hour — clean, professional format. It does the heavy lifting so I can focus on the details."
"Go for it — I'm all for utilizing new tools. Heck, I hesitated about 30 seconds when I traded my slide rule for an HP-32S."
If you train us, do it right
"If there will be training, make it simple, steady and complete. Simple: everyday language, we're not coders. Steady: regular and repeated. Complete: we really understand it before we're turned loose."
To decide together — Meeting 2
What the data puts in front of us
Not conclusions — the open calls the survey sharpens, in the order the numbers press hardest:
- 01Approved tools & accounts — the urgent one.Heavy use is on personal ChatGPT. Do we lead with Copilot (in-tenant on our M365) plus a short approved list, and how fast can we provision company accounts?
- 02Two on-ramps, one policy.Office wants training; field wants clear rules first. Do we split the rollout — guardrails + hands-on examples for the field, enablement for the office?
- 03The 17 non-users.Training + role examples + someone to ask. Who owns that on-ramp, and is it opt-in or expected?
- 04The verify rule.The company's top worry matches ours. Lock the language: AI output is a draft, never a finished product — a person verifies before we rely on it.
- 05Take the skeptics seriously.A real minority worry about dependence and dulled judgment. How does the policy protect the craft, not just enable the tool?
Every chart here is anonymized and drawn from all 55 responses. The raw open-text answers are available for the group if we want to read more in full.