Ask the building.
Every drawing, manual, photo, email, and invoice your buildings produce — A4 holds them all. Ask anything. Watch the agent work in your browser. The answer arrives with the page it came from. Tough cases come back as a comprehensive, auto-generated report in hours.
Ask. Watch. Read. Investigate.
One window. The whole record. The library, the agent, the side-drawer of citations, and the forensic investigations that close the case — in one continuous flow.
One question. One window.
A senior superintendent types a sentence into A4. "What's the warranty status of CH-04 and when was it last serviced?" Or "Pull every invoice from Carrier in 2024 over $5K." Or "Reconstruct five years of replacement history on the generator pair."
The question lands in a chat. The chat routes to one of three modes — Assistant for ad-hoc questions, Claude Code for engineering work, Live Terminal for the kind of investigation that needs a shell. Your team doesn't pick the mode. A4 does, then exposes the choice as a control if you want to override.
You watch the agent type.
A4's terminal is not a chat box pretending to be an agent. It is a containerized terminal running in your browser, streaming the agent's keystrokes over a live connection. You watch the agent type — search the document library, read the manual, query the asset register, write the finding.
Your team has been told for two years that "AI is the future" without ever being shown an agent doing actual work on actual data. This is that demo. You watch the work happen.
The answer slides into view.
When the agent finds something — a manual page, an invoice, a photo, an email thread — those documents do not stay buried in the terminal scrollback. A drawer slides in alongside the terminal showing the curated set; clicking a card opens the document with the page-cited paragraph highlighted.
The agent does the work in the terminal. The result is in the drawer. Nobody has to reconstruct the trail. That library is 271,000+ documents deep on our first customer's tenant, with 2.2 million indexed passages behind it.
When the question outgrows the chat.
When the question is too big for a chat, the Workspace escalates: a phased plan generates, parallel workstreams fan out, every finding cites a page, and a finished PDF lands on the desktop with sections and an evidence appendix.
The forensic. "Reconstruct the parts-replacement history on these five generators." A five-year reconstruction across five engines — 29 events, nine iterations in a day until the ledger balanced. Replaces the $10,000–$30,000 forensic-engineer engagement a building owner hires the day before a refinance, an insurance claim, or a defect dispute.
The field locate. "Find every photo, email, and drawing for extractor fans E1–E14 — an engineer needs locations now." Locations returned. 29 photos, 12 emails, 30 drawings in one PDF, every source reproduced inline so the engineer can read it offline at the asset.
Twenty-three investigations completed on our first customer's tenant.
Ninety-minute hunts in half a second. Multi-week reconstructions in hours. One window.
Every document is read three ways. Every photo, two more.
Your team doesn't upload, OCR, classify, tag, chunk, index, or summarize anything. A4 does — every document, every day — in four steps:
- Read it. Synced from email, SharePoint, an integration, or uploaded. A4 extracts the text from the PDF, scan, photo, or spreadsheet, and breaks it into passages sized so a question lands on the right paragraph — sentence boundaries preserved.
- Index by meaning. A4 finds the passage that says the same thing as the question, even when the two share zero keywords.
- Index every photo two ways. Click a photo of a leaking flange and get the other photos of leaking flanges. Type "leaking flange," see the same result. No tags.
- Extract and link. Vendor, model number, dates, dollar amounts, equipment references — written to structured fields on every document, classified into 65 document types automatically. A4 then draws an edge between every document and every entity it mentions. 34,000+ links on our first customer's tenant. 21,000+ verified by a human.
The customer didn't run an indexing project. roughly 60% of the corpus arrived as email — Outlook and SharePoint auto-sync round the clock, the pipeline does the rest.
The asset register and the maintenance log. In production. Today.
Two cards. Both production-real. Both load-bearing. The Workspace runs on top of these.
Open the asset. The record's already there.
Open chiller CH-04 and the OEM manual is page-cited. The install invoice is one click. The last service photo is in the gallery. The warranty period is computed from the invoice date. The next PM is on the schedule.
Across A4, 34,000+ documents have been linked to entities — equipment, vendors, invoices, investigations — almost all of them auto-discovered by the indexing pipeline; 21,000+ have since been verified by a human. The screen on the demo wasn't seeded for the demo; it's seeded by the same pipeline that runs every night.
Replaces the integration project the buyer keeps deferring. The asset-data fragmentation that costs every CMMS migration its first six months.
Walks like Excel. Barks like Excel.
The facilities team that's logged maintenance in .xlsm files since 2008 isn't going to get retrained on a six-week onboarding. The PM Workbook ingests the spreadsheet they already have — frozen panes, conditional colors, embedded images, the OVERDUE tab, the HOURS tab.
The engineer hits spacebar to stamp the latest equipment-hours reading. The cell paints instantly. If another engineer just edited the same row, the page tells you and refreshes. Every edit is in the audit log.
Replaces the spreadsheet-and-CMMS split your director of engineering is tired of explaining. The version-conflict overwrites you got burned on once.
Three more capabilities.
Book a 30-minute walkthrough this week.
You bring an active question — a defect, an investigation, a document hunt, an asset record you can't reconcile. We open A4 against your data, in your browser, and show you the answer arrive. You leave with the screen-recording, the underlying citations, and a written estimate of what A4 absorbs from your current stack.
If you'd rather skim before you talk, request the capability matrix below. We email it back to you; the lead lands in our inbox. We won't spam you.
Get in touch
Pick what you'd like below — we'll get back to you within a business day.