Foundry / How we work
+ How we work

Four steps, start anywhere.

Each step stands on its own, with a deliverable and a price. Take the next one only when the last has proven its value.

+ The engagement

Every step earns the next one.

Enter anywhere. Each rung has a deliverable, a price, and a decision point before the next.

01
Assess
Free

Start with the free AI maturity assessment: 15 minutes, conversational, scored across seven dimensions. It replaces guesswork with a clear read on where you stand and what to do first.

  • +A scored read on where you stand
  • +The two or three gaps to close first
  • +A report you can share with your team
Start here, free →
02
Discover and blueprint
From $2,500

A paid engagement with our experts. We sit with your team, map your stack and workflows, and find the highest-leverage work. You leave with a starting AI Playbook whether or not you build with us.

  • +A starting AI Playbook, yours to keep
  • +A clear build plan with effort and ROI shape
  • +A price for every next step, before you commit
03
Build
Priced to value

Delivered through the working model you choose: we build in your environment, we build on our infrastructure and make it available to you, or we guide your team as they build. The highest-leverage capability first, proven in production before we extend to the next.

  • +Working software on the stack you already run
  • +Weekly demos, a human at every consequential step
  • +Proven against ground truth before it ships
04
Extend
Priced to you

Keep us on retainer at the level that fits: coaching, ongoing support, an embedded expert, or fully managed. What we build maintains itself and proposes its own improvements; the only question is who approves them. Your team runs it, our expert runs it with you, or we run it for you.

  • +Support and advisory between builds
  • +Education that raises the floor and the ceiling
  • +A partner who already knows your business
We price each build to the value it creates, and you approve every step. Stay with us as deeply as you want, from coaching to a full strategic partner.
+ Who runs it

The build is one decision. Who runs it is another.

Every build ends the same way: working software, and a choice. Pick the level of ownership that fits today, and change it as you grow.

Own it

Your team runs it.

We build with your people and hand over the keys, with the runbook and the training to own it outright. We stay a call away.

Embedded

Our expert runs it with you.

One of our experts joins your team and leads the AI work from the inside: building, shipping, and raising your team's fluency as they go.

Managed

We run it for you.

We operate what we built: the agents, the workflows, the knowledge layer. We watch the quality bar and ship the improvements. You get the results, with the receipts.

Same system underneath, whoever runs it. Most clients move along this dial over time, and nothing locks you in.
+ What it feels like

Your first engagement, phase by phase.

The shape of a typical first engagement, from the free assessment to a working build your team runs.

Phase 01
Setup and discovery
Accounts, access, and a map of the work worth doing. Your security and IT questions answered up front.
Phase 02
First build
The highest-leverage workflow, built with your team in regular working sessions.
Phase 03
Prove it
Recreated against ground truth and measured, until your team trusts the output.
Phase 04
Hand off and decide
Your team runs it. The next build is optional, and priced before you say yes.

Phases, not a fixed timeline. We set the length of each one with you, around your team, your stack, and your pace.

+ What we build

Built to replace the sprawl, not add to it.

Most teams already run a handful of AI tools that never talk to each other: subscriptions nobody fully uses, the same data typed in twice, a workflow that stalls the moment one person is out. Everything below replaces that sprawl instead of adding to it, so the time, budget, and people it frees up go back to the work only your team can do.

+Agentic employees and organizations
Digital teammates that carry real work, from a single role to a whole operating layer, amplified by your experts.
+Automated workflows
Almost any process, end to end, with a human at the moments that matter. One workflow instead of the spreadsheet, the copy-paste, and the three tabs it used to take.
+Connected knowledge layers
One layer of context across your CRM, docs, and data, for your people and your agents, in place of the five disconnected tools that each hold half the picture.
+Evaluation frameworks
Quality you can measure on every run, so trust is earned, not assumed.
+Oversight and governance
Leaders stay in control. Approvals, audit trails, and policy built in.
+Personal agents: your Proxy
A personal agent that knows what you know. It sat in your meetings and read your docs; it answers the routine asks, drafts in your voice, and covers your out-of-office, so that time comes back to you.
+ Five capabilities

Every build above runs on the same five capabilities, live below.

This is a representative run, not a slide.

Agentic platform
Agents that do real work in your core workflows.
Agentic platform
Live
Scheduled run · weekly client report
Fresh data pulled · metrics computed in code, not estimated
Context loaded from the knowledge layer
Draft written in your voice, in your template
Held for review · nothing ships without a person
Approved and delivered · logged for the next run
One workflow, end to end. A human approves the moment that matters.
Meetings
Email
Docs
CRM
Dashboards
Universal knowledge layer
Your people
Your agents
One connected layer of context. Agents and people read from the same truth.
Accuracy
96
Groundedness
92
Freshness
89
Policy compliance
99
Expert-guided checks on every run. The bar at 85 is the pass line, and it rises.
Client deliverable · approved by a person, then sentapproved
Refund over threshold · waiting on approvalpending
Names and PII redacted before any model reads themredacted
Action outside policy · blockedblocked
Every action above, logged with full audit traillogged
Consequential actions wait for a human. Everything is logged.
Sales
3274
Operations
4182
Marketing
2869
Team fluency, before and after enablement. We raise the floor and the ceiling.
Representative run. Client data stays private.

Step one is free, and 15 minutes.