We’re excited to officially launch AI Analytics. Type a plain-English prompt, and it builds an interactive financial report straight from your ERP and project management data, everything from a WIP, to over/under billing, to budget versus actual. No exports, no spreadsheet joins.

It comes with two interfaces to the same data: build persistent, shareable reports in Agave's web portal, or ask ad-hoc questions from inside Claude, ChatGPT, or Microsoft Copilot with Agave MCP.

What AI Analytics does

The problem is familiar: committed cost and change orders live in Procore or Autodesk Build, actuals live in the ERP, and every report that needs both starts with two exports and a spreadsheet, one that's stale by the time it circulates. We built AI Analytics to skip all of that. Here's what it does.

Describe a report, get the report. Type "build a live WIP dashboard with a project filter" or "compare AR invoices to contract value across recent jobs" and AI Analytics builds it in minutes, from live ERP and PM data. It reads through the same connections that power ERP Sync, so there's no warehouse to stand up and no BI consultant on retainer.

Reports are links, not files. Every report persists in Agave. Pin it, share it, and anyone who opens it sees current numbers. Data refreshes daily by default and can be set hourly. View-only readers are unlimited and free.

Iterate by prompt. Wrong calculation? Tell the report what's off and pin the corrected version as the team default. Need it in Excel or PDF? Ask for an export button and it adds one. Want it broken out per project manager? Say so.

Cross-system by default. Because it reads the ERP and the PM system together, it surfaces what neither shows alone: underbillings, billings that don't match contract value, fee erosion on jobs that still look healthy. Click a variance and land on the job cost entry, invoice, or time card behind it.

Read-only and permissioned. AI Analytics can't create, edit, or delete records in your systems. Permissions are role-based, employee and payroll records are excluded at the integration level, and your data is never used to train AI models.

What we learned from a 10-customer pilot

We piloted AI Analytics this spring with ten contractors over four weeks, across Procore, Autodesk Build, Viewpoint Vista, Spectrum, Foundation, CMiC, and Sage. Those ten made about 44,000 requests.

Here's the lesson we didn't expect: there was no single killer use case. We assumed WIP would dominate. Instead, every contractor pointed the tool at a different job. One controller built underbilling reports. One built a weekly sync-health digest. One built portfolio reporting across an entire customer program. One built monthly profit reporting. Several found metrics they hadn't thought to look for — labor per dollar, misclassified jobs, fee erosion on jobs that still looked healthy.

That's why AI Analytics isn't a fixed report library. The reports you need depend on your ERP, your divisions, and who's asking.

Early results

Within weeks of connecting AI Analytics, one contractor found a $1 million underbilling to an owner. It turned up in a cross-system report that hadn't existed before, because building it by hand was never worth the time.

Agave MCP is pulling its weight too. One roofing contractor was re-keying up to a thousand job-cost reports a season from their ERP into their CRM for the CEO's dashboards. Now Claude reads the job-cost data through Agave MCP and writes it into the CRM through the CRM's own connector. Nobody re-keys anything.

Availability

AI Analytics is available today. For existing Agave customers, setup is one or two meetings — about an hour total — to connect your systems, build your first reports, and train the team. The connections are already in place. Report authors are licensed seats; view-only access is unlimited and free, so your whole exec team can read a report without buying anything. And because AI Analytics runs on Agave's own AI capacity, you don't need a ChatGPT or Claude license to build reports.

Rather ask questions inside ChatGPT, Claude, or Microsoft Copilot? Agave MCP exposes the same read-only data to the assistant your team already pays for. And because it sits alongside your other connectors, it can chain steps a portal can't: pull change orders, check email for sub pricing, and draft the chase emails in one pass.

Coming next: scheduled reports that email themselves to each project's PM, uploaded CSVs and PDFs as data sources, more record types (change events, ledger accounts, bank transactions), and controlled data modifications, with confirmations, guardrails, audit trails, and role-based permissions. Reads stay the default; writes will always need a human sign-off.

About Agave

Agave is the AI platform for construction financials, built to work seamlessly with the systems contractors already use. It serves general and specialty contractors, connecting more than 14 ERP and project management systems across cloud, hosted, and on-prem environments. Profitable and backed by Accel and YCombinator, Agave is used by more than 500 contractors across North America today.

‍