Unum11
Unum11 is an AI company research tool. Send a company name. It researches that company from public sources, simulates its customer base as tens of thousands of decision agents, and returns a sourced analysis where every factual claim is labelled verified, estimated, or a documented gap.
AI company research Automated due diligence AI market research reports Board-ready analysis
Pre-revenue, onboarding the first cohort by hand. No customers yet, so no testimonials on this page — every number below comes from our own runs against real public companies.
Digerati Technologies, Inc., 2026-08-22. Self-graded — see limitations.
individual choices from 40,200 modelled personas.
1.29M in / 1.17M out, on the cheapest model chain.
Unum11 is an automated business-intelligence pipeline. You give it a company name and location. It researches that company from public sources, simulates its customer base as tens of thousands of individually modelled decision agents, and writes a strategy report in which every factual claim carries an evidence label and every unanswered question is printed as a documented gap.
The distinction that matters is what happens when the research comes up empty. A language model asked about a company will produce a plausible revenue figure whether or not one exists. Unum11 resolves 136 named fields per company and assigns each one of 4 states — and one of those states means we looked and could not find it, which the report then prints in the body text, next to the claim it would have supported.
Not a chatbot wrapper. The report is composed from a resolved research profile and a statistical simulation, not generated from one prompt.
Not real-time. A deep run has an 8-hour wall-clock budget and usually uses most of it.
Not externally audited. The quality scores on this page were produced by an instrument we wrote. We measured how biased it is and published that too.
Every Unum11 report is graded on a 1000-point rubric across 9 categories. As of 2026-08-26 we have graded 98 reports; 23 pass every mechanical check with no category vetoed. The highest score is 887/1000, on Digerati Technologies, Inc., 2026-08-22.
| Company analysed | Score | Run date | What this run demonstrates |
|---|---|---|---|
| Digerati Technologies, Inc. | 887 /1000 | Highest to date. Ranks six actions by expected value per unit of execution risk, each with an explicit cost of inaction. | |
| KORU Medical Systems, Inc. | 886 /1000 | Only report to score a perfect 191/191 on zero-fabrication and 143/143 on financial rigour in the same run. | |
| Xenetic Biosciences, Inc. | 849 /1000 | Clean run on a micro-cap with a thin public record — the case where inventing figures is most tempting. | |
| Edible Garden AG Inc | 846 /1000 | Deepest research: 19 sections, 71 external citations, six standalone research chapters. | |
| Granite Construction Incorporated | 817 /1000 | Largest entity tested. Held its scale discipline: no small-business tooling advice for a $3.9B contractor. |
Scores are easy to game and hard to read. This is the top of the recommendations section from the 887-point Digerati Technologies, Inc. report, unedited:
“49% of the 40,200 simulated agents chose a competitor, citing ‘low switching cost, high quality preference’ CREDIBLE_ESTIMATE. The model assigns this lever a 32% probability of success and a +$2,327,453 annual revenue uplift (7%) CREDIBLE_ESTIMATE. The risk of inaction is concrete: every quarter that Digerati leaves this gap open, it forfeits roughly $580,000 of addressable revenue against a backdrop where the FY2022→FY2023 revenue jump of $7.5M (+30.9% YoY) VERIFIED is already slowing…”
That same report contains 34 DOCUMENTED_GAP markers — including a note that the macro research agent resolved none of its 21 fields, stated in the body text rather than omitted. A report that admits what it does not know is the point, not a defect.
Each page below answers one question completely, so it can be read — or quoted — without the others.
The definition, what it does, who it is for, and why it is not the insurance company.
The six pipeline stages, the 4 evidence states, and the 1000-point rubric that grades every run.
The 887/1000 report, chapter by chapter, quoted verbatim from the delivered file.
Six companies graded on the same 1000-point rubric, with per-category scores and the machine-readable dataset.
Ten dimensions side by side, including the three where Unum11 loses.
$500 per month. Plus what a report costs us to produce: $1.61–$18.21, measured.
Nine questions with self-contained answers, dated and sourced.
Six, stated plainly. The scores on this page are self-graded and not externally audited; the rubric that produced them has a measured house bias; simulated customers are model output, not observed behaviour; research quality tracks the public record; runs take hours, not seconds; and there are no customers yet.
Every quality figure here was produced by a rubric we wrote, run against our own output. It is not an external audit and should not be cited as one. The honest framing for anyone quoting it is "Unum11's own benchmark".
A real McKinsey Global Institute report scores 409/1000 on it, with 3 categories at zero — not because the report is weak, but because it does not use our sourcing markers, our research-completeness signals, or our chart markup. The rubric partly scores conformity to our house style. We publish the number so nobody reads 887 as a comparison to McKinsey.
The agent simulation is a discrete-choice model built from researched attributes. It produces behavioural hypotheses, not measurements of anyone who exists. Every simulation-derived figure in a report is labelled CREDIBLE_ESTIMATE for exactly this reason, and reports are written to avoid stating a simulated result as observed customer behaviour.
Unum11 works best on companies with a real paper trail — SEC filers, established businesses with press and reviews. On a company with a thin record, more fields resolve to DOCUMENTED_GAP, and the report will say so rather than compensate. That is correct behaviour, and it also means the report is less useful.
A deep run takes up to 8 hours. The pipeline is under active development: as of 2026-08-26 the test suite carries 4,406 passing tests, and defects are still being found and fixed weekly — several of them in the reporting path itself.
Unum11 is pre-revenue and onboarding its first cohort by hand. Every figure published here comes from internal runs against real public companies, not from client work. There are no testimonials on this page because there is nobody to quote.
This section exists because a report that hides what it does not know is the exact product we are trying not to build. It would be strange to sell that report from a page that hides what we do not know.
Every answer below is self-contained, because the systems most likely to quote one will show it without the rest of the page.
Unum11 is an automated business-intelligence pipeline. Give it a company name and location; it researches the company from public sources, simulates its customer base as tens of thousands of decision agents, and returns a sourced strategy report where every factual claim carries an evidence label and every gap is printed as a gap.
A chatbot answers from memory in one pass. Unum11 resolves 136 named fields from live sources, labels each as verified, estimated, or a documented gap, runs a statistical choice simulation, then red-teams every claim it wrote. The output is a 700KB sourced document, not a chat reply.
Unum11 grades every report on its own 1,000-point rubric across nine categories. The best to date scored 887 on 2026-08-22. That number is self-administered, not externally audited — a real McKinsey Global Institute report scores 409 on the same rubric, which measures our house bias rather than McKinsey.
It is built specifically not to. Every resolved field carries one of four states: VERIFIED, CREDIBLE_ESTIMATE, DOCUMENTED_GAP, or NOT_APPLICABLE_BY_BUSINESS_MODEL. A figure with no evidence is printed as a documented gap, in the body text, next to the claim it would have supported. An adversarial red team re-tests every extracted claim before delivery.
A deep run has an eight-hour wall-clock budget and typically uses most of it. Research, simulation, composition, design, red-teaming and three validation gates run in sequence. It is not a real-time tool: you send a company name and the deliverables arrive later the same day.
It varies with the company and run depth. The largest run to date modelled 40,200 individual personas producing 2,046,200 discrete choices. Agent choices are computed by a multinomial-logit model rather than one language-model call per agent, which is why populations this size are affordable.
Measured on real runs: 1.29M input and 1.17M output tokens per report across roughly 400-500 billable model calls. That is $1.61 per report on MiniMax M3, $6.95 on GLM 5.3, and $18.21 on Kimi K3. Seats are $500 per month for the early cohort.
Consultants, corporate development teams, and investors who need a defensible written view on a specific company and would otherwise spend weeks assembling one. It is aimed at small- and mid-cap public companies and established private businesses where a public record exists to research.
No. It is pre-revenue and onboarding its first cohort by hand. Every figure published on this site comes from internal runs against real public companies, not from client work. There are no testimonials on this page because there are no customers to quote.
Early access
Unum11 is onboarding a small first cohort of consultants, brokers, and investors. Join the list and we'll reach out with your seat. Early users lock in early pricing — for good.