CH 03 · Services

Stage 2

Lay the foundation

An AI foundation your company can build on

The foundation is the hard work, and everything after it runs on it. Over 8 to 12 weeks we get the rules out of people's heads, map your workflows, set guardrails, roll out one AI assistant, train your team in waves and clean up the data that matters. Once the guardrails are set, we build custom skills and plugins for the AI assistants your team already uses, so teams can start using AI on real work while the rest of the foundation is built. It starts with discovery, then we build, and you see a working prototype early.

Typical length: 8 to 12 weeks

Start your foundation project.

The floppy disk · made software portable · replaced by the CD, then the cloud

Rewind

What gets in the way

  1. How the business runs lives in people's heads, so AI has nothing to work from.

  2. AI use is individual and informal, with no shared rules or guardrails.

  3. The data that matters is spread across systems and hard to trust.

  4. People have AI tools and no training on how to use them well.

Cassette label image for Stage 2, Lay the foundation
Play

What the foundation project covers

Side A6 tracks
  1. Context captureWe get what your key people know written down: the rules, the exceptions and how decisions get made. Sometimes that means recording an expert at work and turning it into notes.
  2. Workflows and business rulesWe map how the work gets done, with the people who do it, and write down the business rules for your leadership team to approve.
  3. Governance and guardrailsEveryone gets one approved AI assistant and clear rules for using it: which data can go where, AI drafts and people own the result, and one person accountable for it all.
  4. Choosing your AI assistantWe compare assistants on cost, security, admin controls and fit with the tools you already use, then test one in a small pilot with pass or fail criteria set up front.
  5. Rollout and training in wavesThe assistant and its first skills and plugins go out in waves, with training sessions, office hours, a champion in each department and a playbook for each team.
  6. Data and architectureWe clean up the data that matters and set up a governed copy, so nobody queries your live systems. We design the architecture and review it with your IT team.
Fast forward

What you have at the end

Cassette label image for what you have at the end of Stage 2
  1. Workflows & rules, approved

    Your core workflows mapped and your business rules written down, approved by your team and ready for AI to use.

  2. Guardrails in place

    An acceptable use policy, clear tiers for which data AI can touch, and one owner accountable for how AI gets used.

  3. A trained team

    One AI assistant in everyone's hands, set up with admin controls, and people who know how to use it in their own work.

  4. A first set of skills & plugins

    Skills are saved instructions, connections and context so AI does a task the same way every time. Plugins package skills, and sometimes a read-only connection to one tool, so your team installs them in one step. We build the first set for your team's most common tasks and roll it out with training. Your admin governs them, and more get added through advisory support or in stage 3.

  5. A working prototype

    Something real, built on your own workflows and data, to react to before you decide what to build next.

  6. A plan for what's next

    A prioritized list of what to build next, with an architecture design your IT team has already reviewed.

Why StealthX

What makes us different

Cassette label that reads Doers, with five rows of bars, each row in one ink and offset from the last
  1. Mid-market focus

    We work with leadership teams at privately held mid-market companies, $30M to $500M in revenue, and the private equity firms that own companies like them.

  2. We test it first

    We try new AI tools on our own business before we bring them to you.

  3. Invest for the long term

    We care about your long-term success and invest in deeply understanding your business, customers, and employees.

  4. Always custom

    Everything we build is made for how your company runs.

  5. Working prototypes, early

    We get to work fast and show you working ideas early in a project.

Questions

Common questions

How long does the foundation project take?
8 to 12 weeks. It starts with discovery, then we build. You get a status review every week and see a working prototype early in the project.
How do we let people use AI safely?
Give everyone one approved AI assistant with admin controls, clear rules on which data can go where, and skills your admin governs. Connections to your systems often start read-only, so AI can look and can't change anything.
Will the foundation be outdated as the technology moves?
Your written rules, workflows and context stay useful whichever AI tools you use next. They are what every later stage runs on, so the work carries forward as the technology moves.
Do you do data engineering?
No. We don't do data warehousing or data engineering. We work alongside your data partner, agree up front who owns what, and focus on the data that matters for the work you want AI to do.
Can't we just build the one use case?
We build that use case as a module of your AI operating system, on top of the foundation. The next use case reuses the same rules, data and connections, so each one gets faster to build.
Is training included?
Yes. The assistant and its first skills and plugins roll out in waves, with training sessions, office hours, a champion in each department and a playbook for each team.
When does our team start using AI on real work?
Early in the project. Once the guardrails are set, we build custom skills and plugins for the AI assistants your team already uses and train people to use them. Teams can start solving their own problems while the rest of the foundation is built.
Cassette label image for common questions about Stage 2

Start with the foundation

A foundation project of 8 to 12 weeks, with discovery first and a working prototype early on.