For pharma & biotech R&D and life-science consultancies
The landscape review you can defend line by line.
You have three weeks, two hundred documents, and a leadership deck at the end of it. Fiducial reads the documents — including the internal ones — and gives you a report where every claim opens the exact passage it came from. When someone asks where a number came from, you click.
Your documents are never used to train models. Zero-retention processing and security review before anything of yours moves.
Executive summary
Across the 214 documents reviewed, three of the four candidate approaches show consistent directional benefit1, while the fourth is supported only by single-site data that later work did not reproduce2. Two internal memoranda contradict the published record on dosing3; that conflict is unresolved and is flagged rather than smoothed over4.
Apparatus
Each superscript resolves to a highlighted passage in the source document — internal memoranda included.
Hover to inspect · Brown et al., 2020 · arXiv:2005.14165
“175 billion parameters, 10x more than any previous non-sparse language model”
Inspection
Your reviewers audit claims, not the corpus.
Every load-bearing claim opens the page it came from, with the passage lit. Review becomes clicking through the claims that matter rather than re-reading 500 documents.
Use cases
Three reports people actually ask for.
Technology landscape review
You need to tell leadership where a field actually stands. We read the published work and your internal memos together, and show who is doing what — with the evidence behind each statement one click away.
Competitive evidence synthesis
Someone will ask how our candidate compares. You get a side-by-side table your medical and strategy teams can interrogate row by row, with every number traceable to the paper it came from.
Policy evidence report
Your recommendation will be challenged by people who disagree with it. Every line traces to the study behind it, so the argument survives the scrutiny.
The alternative
What this replaces.
| Commissioned report | Analyst by hand | Free AI tools | Evidence Report | |
|---|---|---|---|---|
| Cost | $40–80K+ | Weeks of senior time | “Free” | $1,950 |
| Turnaround | 6–8 weeks | 3–6 weeks | Hours | Days, incl. your review |
| Reads internal docs | Yes, under NDA | Yes | No — open web only | Yes, ≤500 documents |
| Claim-level audit | Footnotes, if asked | Analyst-dependent | None enforced | Every claim traceable |
Figures are typical ranges for commissioned landscape work. Fiducial replaces the reading and assembly, not the strategic judgment — which is why life-science consultancies run it too: the same engine behind your deliverable, produced in days, with senior time spent on the analysis clients actually pay for.
Security & privacy
Built for documents that can’t leak.
No training on your data
Your documents build your report and nothing else. Neither we nor our AI provider train models on them — Anthropic’s commercial terms prohibit training on API content.
One AI provider, and a short vendor list
Generating a report requires a model to read your text, so passages are sent to Anthropic’s API for inference — the only AI provider we use, under terms that prohibit training, with zero-data-retention available for your organisation. The hosting and storage vendors that hold your files are named in our sub-processor list, supplied with the security pack.
Encrypted, and deleted on demand
TLS in transit, encrypted at rest. Delete a project and its documents, embeddings, and graph are removed from active systems immediately, then purged from backups within the retention window. An audit log records what was processed and when.
Honest compliance posture
SOC 2 is on our roadmap, not yet certified — we say so rather than imply otherwise. If your corpus contains PHI, our provider supports HIPAA-ready processing under a BAA. Security questionnaires answered on request.
Pricing
Less than a week of a consultant's time.
Early reports are built by hand: we set up the columns you asked for, review the output ourselves before it reaches you, and record a walkthrough so you can see how it was put together. Unassisted self-serve reports will be priced substantially lower once the pilot programme completes.
Evidence Report — Pilot
$1,950
per report · limited slots
- ≤500 documents, incl. internal PDFs
- Landscape-report template
- Custom extraction columns
- Temporal map + gaps & contradictions
- Scoping call + bespoke extraction schema
- Expert-in-the-loop review before delivery
Systematic Review — Pilot
$2,500
per review, from
- Protocol → search → screening → PRISMA
- OpenAlex + PubMed search
- Human-confirmed screening decisions
- Audit-log export
- Limited pilot slots
Programme
Talk to us
multi-report & recurring
- Multiple reports or recurring monitoring
- Shared extraction schema across reports
- Security review & questionnaire support
- Volume pricing
Queries
What buyers ask.
Who does the verifying?+
Your team does — that’s the point. Every claim links to its source passage, so review means clicking through the load-bearing claims rather than re-reading the corpus.
You send our documents to an LLM — how is that private?+
Generating a report requires a model to read your text, so passages are sent to our one AI sub-processor (Anthropic) for inference. Processing is not training: inference reads your text to produce output, it does not update model weights, and Anthropic’s commercial terms prohibit training on API content. Zero-data-retention — where prompts and responses are not stored after the response returns — can be enabled for your organisation before any of your documents move. No other AI provider sees your content; the hosting and storage vendors that hold your files are named in our sub-processor list.
How do you handle confidential documents?+
Encrypted in transit and at rest, deleted on demand with an audit log, and never used for training. We’ll complete your security questionnaire and enable zero-retention processing before any documents move.
What’s the turnaround?+
Ingestion and drafting typically complete within a day; the calendar time is your team’s review. Most reports are final within a week of corpus handoff.
What formats do we receive?+
DOCX and PDF using the landscape-report template, plus the live workspace where reviewers click from any claim to its source. LaTeX on request.
How does a pilot work?+
Send a sample of your corpus and the columns you care about — a short form, not a meeting. We run the template against it and send back the report plus a walkthrough video. Everything is async; no call needed. If the output isn’t useful, you don’t proceed.
Bring the documents. Leave with a report that holds up.
Send a few documents and the columns you care about. You get a sample report back, with a short walkthrough video — no call required.