You are paying for AI. Metricept tells you what it is actually returning.
Vendors report adoption. Consultants estimate savings. Metricept grades every claim by the evidence behind it, counts only the time that was really put to use, and gives each AI investment a verdict: expand, maintain, investigate, or stop.
The numbers in your AI business case were written by people who wanted the sale.
Nobody can tie AI spend to profit
Fewer than one in three leaders can connect AI value to the P&L. Renewals get approved on faith.
Most licences sit idle
In Microsoft's own trial, only about four in ten workers became regular users. Business cases assume ten in ten.
"Hours saved" go nowhere
Freed time is counted as savings. In most organisations less than half of it is redeployed to paid work.
Vendors grade themselves
Every platform now ships an ROI dashboard for its own product. None of them is independent, and none deducts risk or lost quality.
Three steps. One verdict per AI investment.
1 · We collect what you already have
Licence invoices, the usage export from the vendor console, before-and-after numbers from the system that tracks the work. No prompts, no system access.
2 · The engine grades every claim
Each benefit gets a trust level from TL0 (someone said so) to TL4 (finance confirmed it). Only redeployed time counts. Lost quality and risk are deducted. Same inputs, same answer, every time.
3 · You get a decision, not a dashboard
A signed report per initiative: what is proven, what is supported, what is just a claim, a score, a verdict, and the one record that would raise the trust level next.
Put in the claim someone made about your AI tool. See what survives.
Same engine as the report. Runs in your browser; nothing is sent anywhere.
What Metricept tells you that a vendor's ROI calculator won't
Every number shows how much to trust it
Each figure comes with a trust level and the records behind it. If a vendor claims $180K and the evidence supports $36K, the board sees both and knows why they differ.
Time saved only counts when it's reused
Freeing five hours a week is not a saving until those hours go somewhere. Metricept measures how much freed time is actually redeployed. In most organisations it's under half — and that one number decides whether the tool pays for itself.
Faster is only better if quality holds
Output that is 30% faster with 20% more rework is a 4% gain, not 30%. Metricept prices every speed gain against its error rate, and charges lost quality as a cost.
Risk is deducted, not footnoted
Sensitive data in a public model, invented citations in a submitted proposal, dependence on a single vendor — each is priced by likelihood and impact and taken off the return before it's reported.
Every initiative gets a verdict
Expand, maintain, investigate, or restructure — each based on a score and a threshold, not a feeling. Renewal conversations start from the number, not the demo.
You know what to fix next
For each initiative, the report names the single record that would raise its trust level and the person who owns it. The measurement gets better between assessments, not only during them.
Why Metricept, and not the free dashboard.
Independent
We sell no seats and no models. A vendor cannot credibly grade its own return. We can.
Evidence-graded
Six trust levels with fixed acceptance rules. A bigger claim cannot buy a higher score.
Counts real time only
The only method that separates hours freed from hours redeployed — the number vendors never report about themselves.
Safe for regulated work
No prompt content, ever. Runs inside a CUI enclave or your own cloud. Built for defense, healthcare and government.
Research-driven. Metricept is built on a systematic review of the published evidence on what AI actually does to productivity, quality and cost in real companies — not on vendor benchmarks. Every rule in the engine traces back to a finding. We keep the details in the method statement that ships with every report.
Scientists and researchers who got tired of untraceable numbers.
Potomac AI Labs is an applied AI lab in Washington, D.C. We build working systems from research — in automation, cybersecurity and decision-making. Metricept exists because we watched good companies renew AI contracts on figures nobody could defend. Dr. Beza Lefebo, Dr. John Kuk, Dr. John Kramarczyk.
Our AI platform already gives us an ROI dashboard. Why pay for this?
Because it was produced by the company that wants the renewal, and it reports only that company's product. Metricept grades every tool on one scale, counts only redeployed time, deducts risk, and says how much each number can be trusted. The free dashboard is the input. Metricept is the audit of it.
Do you need access to our systems, or our prompts?
Neither. You send exports — invoices, the usage report from the admin console, before-and-after metrics. We never request or read prompt content.
What if we have no real evidence yet?
Then the report says so, precisely, and lists the one record that would change it. Many first assessments are mostly a to-do list. That is the point.
What does it cost and how long does it take?
A fixed fee for one to three AI initiatives over three to four weeks — priced to be an easy decision next to a single renewal. Use the form and we quote on the call.
Bring one AI tool and the number someone attached to it.
Mid-sized companies, defense and engineering contractors, PE-backed portfolios. If you hold a CMMC or CUI obligation, risk is weighted accordingly.