Skip to content

Metis — Internal Knowledge Search for SMBs

Your team accumulates documents over years: equipment manuals, prior projects, SOPs, vendor contracts, change orders, safety procedures, codes. When someone needs an answer that's already in those documents, they spend 20 minutes hunting through file shares — or they recreate work that's been done before. Metis fixes that.

Last updated: 2026-05-14

What it does

Ask a question in plain English. Get an answer with the source document quoted.

Q: What's the start-up procedure for the chiller at Hill Crest?
A: [Source quote from "HillCrest_O&M_Manual_2023.pdf", page 47]
   "Open shutoff valves V-1 through V-4, verify reservoir level
   between 60-75%, set controller to AUTO, allow 4-minute warm-up
   cycle before enabling chilled-water pump..."
   [+ 3 more relevant excerpts from related documents]

Three properties that matter:

  1. Every answer cites a verbatim quote from one of your documents. If the system can't find the answer in your documents, it says so — it doesn't make one up.
  2. Your documents stay yours. Stored in an isolated AWS S3 bucket. Encrypted. The underlying AI models don't train on your data.
  3. Your team learns from itself. Estimators, PMs, field techs all query the same accumulated knowledge — your firm's institutional memory becomes searchable.

Who it fits

Companies that:

  • Have 50-2,000 employees and lots of accumulated documentation
  • Run operations-heavy work (contracting, manufacturing, services, distribution, professional services)
  • Lose meaningful time to "where's that document?" / "how did we do this last time?" / "what did we promise the client on the {project name} job?"
  • Have field or distributed teams who need answers fast and can't always reach senior staff

Industries that fit naturally: mechanical / electrical / HVAC contracting, specialty trades, manufacturing operations, fleet maintenance, building operations, technical services, professional-services firms.

Who it doesn't fit

  • Companies with everything in their head (truly small, single-location, single-decade-of-employees)
  • Companies where the document pain is "we don't have the documents" (Metis searches what you have; it doesn't create what you don't)
  • Customer-facing search applications (this is for your own team, not your customers' search)
  • Regulated-industry compliance audits — that's a separate product (AuditForge)

What's in scope

| Document types | PDFs (equipment manuals, contracts, prior bids, drawings with text), DOCX (SOPs, reports, proposals), CSV (job logs, equipment registers, project lists), Excel (cost histories, vendor lists), emails (Outlook PST / Gmail Takeout), TXT/MD | | Use cases | "How did we do X last time?" / "What's the procedure for Y?" / "What did we promise on the Z project?" / "What's the spec for equipment model A?" / "Which jobs had similar scope to the new bid?" | | Languages | English (additional languages on request) | | Search style | Natural-language questions, not keyword search |

What we connect to

  • Drag-and-drop upload — out of the box
  • CSV bulk import — out of the box (job logs, equipment registers, ticket history)
  • Shared drives (Google Drive, SharePoint, OneDrive, Box, Dropbox) — connector built on request, scoped to your environment, typically within 1 week
  • CMMS / job-management systems (ServiceTitan, FieldEdge, Maintenance Connection, eMaint, Limble, etc.) — connector built on request after we know which system you use
  • Watched-folder sync script — quick-start option for pilots; runs on one machine, auto-uploads new files

See docs.base2ml.com/auditforge/data-sources/ for the full list and how integration scoping works.

Pricing

Pilot — free (60-90 days)

We're looking for one or two friendly customers to validate Metis on real SMB corpora. The pilot includes:

  • Up to 500 documents
  • 5 named users
  • Live corpus ingestion + weekly 30-min check-ins
  • Honest end-of-pilot review
  • Permission for an approved case study

You commit ~2 hours per user during the pilot. We commit ~6 hours of founder time. No fee.

After pilot — month-to-month

Tier Monthly Includes
Basic $400 Single corpus, up to 10 users
Standard $800 Single corpus, unlimited users, monthly corpus refreshes
Annual $600/mo ($7,200/yr paid up-front) Standard features, 25% discount

Month-to-month cancellation. No multi-year lock-in.

What you'd ask a current customer to verify

This is a newer product line at Base2ML. Today there's no published reference customer in the SMB-internal-knowledge space — Metis has been used for related work (audit-firm document review under our AuditForge product) but the SMB use case is one we're validating with the first 1-2 paying customers.

That means today's offer is structurally favorable to early customers — you get founder attention, generous pilot terms, locked-in pricing, and case study approval rights. The trade is that you're an early customer, with all the upside and risk that comes with that.

If first-customer status is a disqualifier for you, the right answer is "let's revisit in 6 months." We won't push.

Honest disclosure

  • No SOC 2 report yet. Substantive controls (encryption, access control, audit logging, isolated buckets) are in place; auditor-signed Type 1 report is on the post-revenue roadmap. For non-regulated SMBs this gap is typically tolerable.
  • No pen test report yet. Same timeline.
  • One-person operation today. Founder responds to engagement-level questions within business hours; there is no 24/7 support tier.
  • Founded 2026 in Pittsburgh, PA. Sole-founder bootstrap so far. Not vaporware (production-deployed, real users have used the underlying engine), but not enterprise-scale-tested either.

What's next

Email chris@base2ml.com with a one-sentence description of what your team currently struggles to find. We'll book a 30-minute call, dig into whether Metis fits, and (if yes) scope a pilot.