AI Systems Assessments

Understand what's slowing your team down, whether your AI tools are helping, and what needs to change before you invest in another solution.

An AI systems assessment from charmthirteen examines whether your organization has the tools, workflows, data, governance and internal skills to use AI productively. It follows Discover, Assess, Recommend, includes hands-on inspection of your systems and data, and ends with clear priorities for what to keep, change, automate, govern or build.

Discover. Assess. Recommend.

Every assessment moves through the same three stages. We discover how the work happens today, assess it against six dimensions, and recommend what to do next, in priority order and in a sequence your team can realistically carry out. Implementation and training are available afterward as follow-on services, and you're never required to buy a build to get a useful assessment.

Assessments include hands-on inspection of your systems and data. We look at the tools, the workflows and the information your team actually uses, because a survey can't tell you what a spreadsheet, an approval queue, or a shared drive will.

We check whether AI can do the task, and then whether the system you'd end up with would be reliable, governable, maintainable, and useful to the people doing the work.

From assessment to implementation. charmthirteen can help design, build, train, and document the recommended solution when the client wants support.

AI Adoption & Enablement

After the assessment, you can bring charmthirteen in to carry out the roadmap with your team:

• Claude Skills, Projects, and reusable workflows
• Tool configuration and troubleshooting
• Department-specific training
• AI office hours
• Internal champions and peer mentoring
• Documentation and handoff
• Ongoing evaluation and improvement

Enablement runs as a fixed-scope project or a time-limited advisory arrangement. It doesn't turn into a retainer unless you ask for one.

Six places AI adoption quietly breaks.

These are six dimensions, not six products. Every package looks across them, and the package decides how far into each one we go.

Data & Semantic Readiness

AI can only work with what your organization has written down, connected, and named consistently. Most readiness checks skip that layer, and it's often the reason an AI tool produces confident answers nobody trusts.

What we examine:

• Data inventory and lineage: where relevant information originates, and how it moves between systems.
• Terminology and definitions: whether different teams use the same terms to mean different things.
• Workflow documentation: whether business rules and approval steps are written down, or exist only in employees' heads.
• Data mapping: whether information can be reliably connected across systems.
• Knowledge architecture: whether a shared semantic model would improve how agents retrieve and interpret your organization's context.

Where knowledge graphs fit

Some teams need better documentation or a searchable knowledge base. Others may genuinely need an ontology, graph-based retrieval, or a richer context layer. A knowledge graph is one possible recommendation when we find fragmented relationships, definitions, or institutional knowledge, and the assessment decides which solution is justified. It is never a standard deliverable.

See a reference architecture from prior work →Reference architecture

How we investigate: the Orbit Framework.

Orbit is charmthirteen's discovery method, and in an assessment it becomes the structure of the investigation itself. The package sets how deep each step goes.

1. 01

   Pre-Orbit + Entry

   Establish your problem, inventory the context available, and identify what needs investigation.

2. 02

   Mapping

   Interview employees, document workflows, identify tools, trace how data moves, and map who is responsible for what.

3. 03

   Gravity

   Quantify what the inefficiency costs, examine root causes, assess risk, and decide whether there's a viable business case.

4. 04

   Circles Out

   Document findings, evidence gaps, unresolved dependencies, and what still needs validation.

5. 05

   Trajectory

   Recommend what to stop, retain, improve, buy, automate, or build, with priorities and a realistic implementation sequence.

Value you can measure, and value you can see.

Some benefits show up as dollars. Others show up as fewer errors, cleaner ownership, or a control that didn't exist last quarter. We measure each one the way it actually shows up.

Safe deployment and accountability show up as observable controls and reduced risk. No assessment can promise an organization zero risk, and this one won't pretend to.

AI productivity value estimator

Illustrative model for a workflow assessment, using adjustable assumptions.