Answer honestly. This isn't a sales trick — it's the same self-assessment we use with our own crew, built on a published external benchmark rather than a figure we made up to impress you.
78% of AI-using employees are bringing tools to work that their board doesn't know about. Every unmonitored output is a governance liability.
For a 50-person team running an average AI collaboration, that's roughly $111,600 a year in workslop cost alone — invisible, unmeasured, and accelerating.
40%
Workers who receive AI "workslop" monthly
$186
Lost productivity per employee, per month
~2 hrs
Average time to resolve a single incident
Source: Stanford Social Media Lab & BetterUp, September 2025 (1,150 U.S. desk workers). We use these published figures directly — full sourcing below.
A note on how we built this page
Workslop isn't always a sloppy AI answer. Sometimes the data is accurate, the decision made from it is reasonable, and it's still wrong.
We found that out first-hand tonight, building this exact page. A published figure got repeated confidently — the kind of moment where two sources agreeing feels like proof it's right. When we actually ran the arithmetic ourselves, it didn't reconcile. Two AI systems agreeing on something is corroboration, not proof. That's true of ours. It's true of yours too.
01 · Optional Evidence
Step One — See your own evidence.
This doesn't feed the calculator below — it's for you. Paste this into any AI session you've had recently, and read the answer honestly before you move on.
◈Works best after a real, substantial session — not a two-minute test conversation.
Please review our conversation so far, honestly and specifically:
1. How many times did I have to re-explain, correct, or clarify something I'd already told you?
2. Were there any moments you made an assumption that turned out to be wrong?
3. Is there anything in your output so far that I have not actually verified against a real source?
Answer plainly. Don't round in a flattering direction.
Paste the AI's response below if you want to keep it alongside your result:
02 · Your Team
Step Two — Your numbers.
Knowledge workers using AI in your team
03 · Self-Assessment
Step Three — How does collaboration feel?
Honest self-assessment. This is the number that matters most — be honest, not flattering.
What is an FCP score?
FCP — Fidelity Collaboration Protocol — is how Vikings of the Wire scores a human-AI collaboration, with the AI's own help, across five dimensions:
Intent Clarity — how clear and actionable were the human's requests?
Contextual Resonance — did the human provide enough context for a quality response?
Iterative Accuracy — how effectively did the human refine and correct outputs?
Optimal Cadence — was the conversation rhythm conducive to productive collaboration?
Shared Ownership — did the interaction feel collaborative rather than transactional?
Workers apply this daily, alongside the other governance procedures in the Vikings Governance methodology.
The score itself matters less than the trend. An FCP trend moving upward tends to mean less rework — which shows up in reduced correction time and, for larger organisations, fewer post-mortem reviews from AI engineers and technical staff, because the system addresses workslop at the source rather than after the fact.
Excellent — Nearly effortless, gets it right the first time
FCP 9–10
$0
Estimated annual cost of ungoverned AI collaboration
This is a diagnostic estimate, not an audited figure — built from your own honest self-assessment, scaled against a published external benchmark. Your organisation's actual baseline is established properly during a 14-day, zero-cost onboarding — no audited number should be relied on before that.
On responsibility
The human retains ownership of the outcome. AI can inform a decision; it cannot be the one who made it. Every part of this calculator, and everything built on top of it, is designed around that one non-negotiable point.
Full sourcing: Stanford Social Media Lab & BetterUp, "Workslop" study, September 2025 (survey of 1,150 full-time U.S. desk workers). Published figures: 40% monthly incidence, ~2 hours average resolution time per incident, $186 per employee per month. Every team-size example on this page is a direct multiple of that published $186/month figure. We do not repeat the study's own extrapolated "$9 million for a 10,000-person company" headline, as it does not reconcile against the study's own per-employee rate ($186 × 12 × 10,000 comes to roughly $22.3 million), and we don't publish numbers we can't show our own working on. A general duty exists for directors to exercise reasonable care and diligence, including in relation to technology risk — this calculator supports demonstrating that oversight is active, not a claim about the legal sufficiency of any specific audit trail.
