A third Anthropic certification
TryfanTech is now certified in Anthropic's AI Fluency for Small Businesses — a programme built specifically for the kind of businesses we work with every day. It's our third Anthropic certification, joining Claude with Amazon Bedrock (April 2026) and Claude Code in Action (March 2026).
The three cover the three sides of how we use Claude in client work. Claude Code in Action is the developer side: the agent SDK, MCP connectors, and the patterns behind building AI workflows. Claude with Amazon Bedrock is the deployment side: running Claude on enterprise infrastructure, with the IAM, networking and observability that come with it. AI Fluency for Small Businesses is the human side: how a small business actually decides what to delegate to AI, how to brief it, how to evaluate the output, and how to stay accountable for what goes out the door.
This post is about the third one — because the framework behind it is the most useful thing we've come across for getting a small team using AI consistently.
What "AI fluency" actually means
Anthropic's definition is precise: AI fluency is the conscious ability to collaborate with AI in ways that are efficient, effective, ethical, and safe. Four words, each pulling weight. Strip any of them out and you have something else — productivity theatre, prompt-engineering trivia, or an unmonitored liability.
The 4D Framework operationalises that definition into four interdependent competencies. They form a chain — weakness at any link breaks the practice. Brilliant prompts (Description) cannot save a bad delegation decision. Accurate outputs (Discernment) do not matter without follow-through (Diligence). You need all four, every time.
The four competencies
Delegation — should AI be doing this at all? The first decision, and the one most often skipped. Reversible, low-cost, language-shaped work is safe to delegate. Customer-facing, legally exposed, financially material work is harder. If you can't verify the output cheaply, the delegation isn't paying off. Delegation is the gate; get it right and the rest of the framework has something to act on.
Description — how do you brief the AI? What the industry calls "prompt engineering" is, at root, professional communication. A strong description includes role, context, task, constraints, examples, and success criteria. Anyone who can brief a colleague well can write effective prompts — it's a transferable skill, not a technical specialism, which means anyone on the team can develop it.
Discernment — is the output any good? Critical evaluation. Accurate? Specific to your business or generic? Complete? Overstated? Confabulated? LLMs are optimised to produce text that reads fluent, confident, and contextually appropriate. They are not optimised for accuracy. Confidence in tone is not correctness in fact. Discernment is the muscle that catches the gap.
Diligence — did we handle it responsibly? Data handling, verification, attribution, sign-off. A named human owns every output that leaves the business. "The AI made an error" is never a defence — it is your output, your customer, your liability.
The two loops
Where the framework stops being a vocabulary and becomes a workflow.
Description ↔ Discernment — the quality loop. Brief → output → critique → re-brief → better output. Iterative, not one-shot. Treating AI as a one-shot tool ("type a prompt, accept what you get") leaves most of the value on the table.
Delegation ↔ Diligence — the responsibility loop. "Should we be using AI here, and did we handle it right?" The loop that keeps you out of trouble. Most AI horror stories in business — leaked customer data, fabricated citations sent to clients, hiring decisions made on confabulated CVs — are a failure somewhere on this loop.
The three artefacts a fluent business walks away with
The framework points every small business at three reusable artefacts. These are the practical output — the things a fluent operator actually has on file.
A business briefing sheet. A one-page summary of your business — mission, customers, services, your role, daily pain points, values, what you would never compromise on. You paste it at the top of every new AI session. Most generic AI output is generic because the AI has no context about your business. The briefing sheet fixes that in one paste.
An AI workflow document. One per recurring task. Trigger, inputs, what is AI-led vs human-led, the prompt template, what you specifically verify before accepting the output, where the result goes, who signs off. Hand it to a new hire on day one. The opposite of ad-hoc use — written, repeatable, auditable.
An AI use policy. Ten lines. What AI is used for, which tools are approved, what data does not go into them, who verifies, who signs off, when customers are told, what is never delegated. Short on purpose. Most small businesses use AI without one, which means decisions are implicit and inconsistent.
If your team produces those three artefacts and uses them, you are already ahead of most SMBs on AI adoption.
What this changes for our clients
Our AI training workshops now use the 4D Framework as their backbone. The vocabulary your team learns in our workshop is the same vocabulary Anthropic publishes against — which means the framework keeps working long after the workshop ends.
Our custom AI development work is shaped by the same framework. Every workflow we build comes with Delegation decisions written down, Description templates checked in, Discernment checks defined upfront, and Diligence — verification, data handling, sign-off — built in from day one. The framework is not just for end users; it is how we engineer.
The certificate is the headline. The framework is the substance. The work is where both come together.
Bringing the 4D Framework to UK small businesses
If your team is using AI ad-hoc — different people, different prompts, no shared briefing sheet, no written policy — the 4D Framework is the cheapest way to get the practice on a footing. We run AI training for UK small businesses built on the framework, with examples drawn from the industries our clients actually work in: hospitality, retail, professional services, logistics, trades, and creative agencies.
A half-day workshop usually gets a non-technical team from "we've heard of ChatGPT" to running their own briefing sheet, an AI workflow document, and a 10-line AI use policy. Full-day sessions go further, with real project work in the second half. Custom programmes shape the examples around your sector.
See the workshop options, or start a conversation about what your business needs. First call is free, and we'll tell you honestly whether a workshop, a custom AI build, or none of the above is the right answer for you.