Before You Start¶

Forbidden Questions

The Fundamental Questions¶

1. Why AI?¶

The forbidden version: "Is AI actually the right solution, or are we using it because it's fashionable?"

What you'll hear instead: - "AI is the future" - "We need to modernize" - "Other agencies are doing it" - "The Minister wants innovation"

What to probe: - What specific capability does AI provide that other approaches don't? - Have we genuinely evaluated non-AI alternatives? - Is the problem well-defined enough for AI to solve? - Are we solving a problem or implementing a technology?

The uncomfortable answer you might find: Sometimes the answer is "because someone already decided we're doing AI." That's not a reason—it's a constraint. Name it honestly.

2. Who actually wants this?¶

The forbidden version: "Is there genuine demand, or are we pushing a solution onto people who didn't ask for it?"

What you'll hear instead: - "Stakeholders have been consulted" - "This aligns with strategic priorities" - "Users will love it once they see it"

What to probe: - Who specifically asked for this capability? - What evidence do we have of genuine demand (not manufactured consent)? - Have we talked to the people who will actually use it daily? - What are users currently doing, and why would they change?

The uncomfortable answer you might find: Often, AI projects are supply-driven (we have AI capability, let's find uses) rather than demand-driven (we have a problem, AI might help). Supply-driven projects have much higher failure rates.

3. What happens if we don't do this?¶

The forbidden version: "What's the actual cost of doing nothing, and is anyone honest about it?"

What you'll hear instead: - "We'll fall behind" - "The current situation is unsustainable" - "We have no choice"

What to probe: - What specifically will happen if we don't proceed? - Is the status quo actually failing, or just imperfect? - Are there simpler interventions that could address the problem? - What's the honest opportunity cost of this investment?

The uncomfortable answer you might find: Sometimes "do nothing" is actually a viable option. The status quo, while imperfect, may be better than the risks of change. This is heresy in project-land, but sometimes true.

4. Who decided the scope?¶

The forbidden version: "Did scope emerge from analysis, or was it predetermined by politics/budget/timeline?"

What you'll hear instead: - "The scope was carefully considered" - "We've balanced ambition with achievability" - "This is Phase 1"

What to probe: - Was scope set before or after understanding the problem? - What was excluded, and why? - Who made the scoping decisions, and what were their incentives? - If the scope was set by constraints rather than analysis, can we admit that?

The uncomfortable answer you might find: Most project scopes are political artifacts—shaped by what can be funded, what can be announced, what fits the timeline. This isn't inherently bad, but pretending scope emerged from pure analysis creates problems later.

5. What's the real timeline?¶

The forbidden version: "Is this timeline based on technical reality or political necessity?"

What you'll hear instead: - "The timeline is aggressive but achievable" - "We've built in contingency" - "The team is committed"

What to probe: - Who set the timeline, and what was their basis? - Have the people doing the work validated it? - What similar projects have achieved this timeline? - What happens if we miss it?

The uncomfortable answer you might find: Many timelines are reverse-engineered from announcement dates, budget cycles, or political calendars. The technical team then has to pretend it's possible. Everyone knows it's fiction, but no one says so.

6. What's the real budget?¶

The forbidden version: "Is this budget based on actual costs, or on what we thought we could get approved?"

What you'll hear instead: - "The budget is adequate for scope" - "We have contingency" - "We can seek additional funding if needed"

What to probe: - How was the budget estimated? - What's excluded (change management? maintenance? contingency for unknowns?) - What happens if we run over? - What's the full lifecycle cost, not just implementation?

The uncomfortable answer you might find: Many budgets are shaped to fit approval thresholds, then supplemented later. The initial budget is a negotiating position, not an estimate. Everyone knows this; few admit it.

The Questions About Assumptions¶

7. What are we assuming about the data?¶

The forbidden version: "Has anyone actually looked at the data we're planning to use?"

What you'll hear instead: - "Data is available" - "We'll assess data quality in the design phase" - "Data owners have committed to providing access"

What to probe: - Has anyone tried to use this data for this purpose before? - What's the actual state of data quality—not the reported state? - Do we have access, or do we have a promise of access? - What if the data isn't what we think it is?

The uncomfortable answer you might find: Data is almost never as ready as people claim. "Available" means "exists somewhere." "Accessible" means "someone might give it to us." "Quality" is someone's optimistic assessment. Build in time to discover reality.

8. What are we assuming about the technology?¶

The forbidden version: "Has this AI approach actually worked for this kind of problem at this scale in government?"

What you'll hear instead: - "The technology is proven" - "Leading companies use it" - "The vendor has references"

What to probe: - Where specifically has this worked in a similar context? - What's different about our situation? - Can we talk to actual users, not vendor-selected references? - What happens if the technology doesn't perform as expected?

The uncomfortable answer you might find: "Proven" often means "worked in a demo" or "worked for a different problem" or "worked with better data." Government contexts are different from vendor showcases.

9. What are we assuming about the people?¶

The forbidden version: "Will the people affected by this actually use it, accept it, or fight it?"

What you'll hear instead: - "Change management is part of the plan" - "Users have been consulted" - "There will be training"

What to probe: - What do the people actually doing the work think? - What's their incentive to adopt this? - What happens to their jobs? - Have we asked them what they need, or told them what they're getting?

The uncomfortable answer you might find: People whose jobs are threatened, workflows disrupted, or expertise devalued will resist. "Change management" often means "convincing people to accept what we've already decided." That's not the same as genuine engagement.

10. What are we assuming about ourselves?¶

The forbidden version: "Does this organization actually have the capability to do this?"

What you'll hear instead: - "We have a strong team" - "We'll upskill as needed" - "We can bring in expertise"

What to probe: - Do we have the technical skills? - Do we have the organizational maturity? - Do we have the political capital? - Have we done something like this before—honestly assessed?

The uncomfortable answer you might find: Organizations often overestimate their capability, especially for novel projects. Past success in different domains doesn't guarantee success here. Honest capability assessment is uncomfortable but essential.

The Questions About Failure¶

11. What would failure look like?¶

The forbidden version: "If this project is going to fail, how will it fail?"

What you'll hear instead: - "We're focused on success" - "We've mitigated the risks" - "Let's not be negative"

What to probe: - What are the most likely failure modes? - What would cause us to stop? - What does "too late to recover" look like? - How would we know we're failing before it's obvious?

The uncomfortable answer you might find: Most projects don't plan for failure, so they don't recognize it until it's undeniable. By then, sunk costs make stopping politically impossible. Define failure early so you can recognize it.

12. Who will be blamed if this fails?¶

The forbidden version: "When this goes wrong, whose career is on the line?"

What you'll hear instead: - "We're all accountable" - "The steering committee provides oversight" - "Success is a team effort"

What to probe: - Who specifically is accountable—named individual? - Do they have authority matching their accountability? - What happens to them personally if this fails? - Are the people with accountability the ones making decisions?

The uncomfortable answer you might find: Accountability is often diffuse (everyone) or detached (someone who can't influence outcomes). When failure comes, either no one is accountable, or someone is scapegoated. Neither is healthy.

13. What's our exit strategy?¶

The forbidden version: "If we need to stop this, can we?"

What you'll hear instead: - "We're committed to success" - "We'll iterate and improve" - "There's no Plan B"

What to probe: - At what point would we stop? - Who has authority to stop? - What would it cost to stop at various stages? - Is stopping actually possible, or are we committed regardless?

The uncomfortable answer you might find: Many projects have no exit strategy. Once started, political and sunk-cost pressure makes stopping almost impossible. If you can't answer "how would we stop?" before you start, you're not in control.

The Questions About Honesty¶

14. What aren't we saying?¶

The forbidden version: "What does everyone know but no one is willing to put in writing?"

What you'll hear instead: - [Silence]

What to probe: - What concerns have been raised privately but not formally? - What's the gap between the official story and the corridor conversation? - What would the skeptics say if they felt safe to speak? - What would a journalist find if they investigated?

The uncomfortable answer you might find: There's almost always a gap between the official narrative and what people actually believe. The gap is information. What's being suppressed, and why?

15. Who benefits from us not asking these questions?¶

The forbidden version: "Whose interests are served by us proceeding without scrutiny?"

What you'll hear instead: - "We've done our due diligence" - "The business case is solid" - "We can't delay for more analysis"

What to probe: - Who gets something if we proceed—budget, profile, career advancement? - Who loses something if we stop—investment, reputation, timeline? - Are the people pushing hardest the ones with most to gain? - Is there pressure to not ask hard questions?

The uncomfortable answer you might find: Projects create constituencies with interests in continuation. Vendors want contracts. Sponsors want wins. Teams want jobs. These interests aren't wrong, but they create pressure to proceed even when proceeding is unwise.

How to Ask These Questions¶

Create Space¶

• Schedule a specific session for "hard questions"
• Invite a devil's advocate with explicit permission
• Use anonymous input mechanisms
• Bring in an outsider with no stake

Frame Constructively¶

• "I want this to succeed, so I need to understand..."
• "Help me address the concern that..."
• "What would we tell an auditor who asked..."
• "If this were a case study in five years, what would we wish we'd asked?"

Document Responses¶

• Record that questions were asked
• Record what answers were given
• Note who was present
• Keep this documentation safe

Accept Incomplete Answers¶

• Sometimes "we don't know" is the honest answer
• Distinguish "we don't know but we're finding out" from "we don't know and we're proceeding anyway"
• Some uncertainty is acceptable; unexamined uncertainty is not

The Permission Slip¶

You have permission to ask these questions before a project starts.

If asking them slows things down, that's information about whether the project should be moving fast.

If asking them creates discomfort, that's information about what's being avoided.

If asking them ends your involvement, that's information about whether you wanted to be involved.

The questions exist whether or not you ask them. Asking them just means you'll know the answers before the consequences arrive.

"The question you're afraid to ask is the one whose answer you most need to know."