Lessons

Most data presentations fail before a chart is drawn, because they are organised around what the analyst found rather than what the audience must decide. This lesson covers how to identify the actual decision, why the analysis order is the wrong presentation order, and what an audience needs to act.

Every funder asks for impact and few organisations can afford real evaluation. This lesson covers what you can honestly claim, the biases that make feedback flattering, where AI genuinely helps with qualitative data at scale, and why generating impact claims is the one thing that ends an organisation.

Nonprofits carry the workload of larger organisations with a fraction of the staff. This lesson covers where tooling genuinely relieves that: supporter communications, donor operations, reporting back to funders, and the free-tier trap that puts beneficiary data somewhere it should not be.

Grant applications look like writing tasks and are mostly evidence and fit problems. This lesson covers what a reviewer is really assessing, why generated applications fail in a specific and detectable way, and how to use tooling on the parts that genuinely repeat without producing an application that says nothing.

Assistants sit where authority, access and urgency meet, which is exactly what impersonation fraud exploits. This lesson covers why the role is targeted, what changed when voice and video became cheap to fake, and the verification habits that hold up when a convincing executive is asking you to move quickly.

The concrete workflows: triaging an inbox that is not yours, building briefing packs, handling travel and what happens when it breaks, and producing documents. What each needs as input, where each fails, and what has to stay with a person.

Executive assistance looks like a set of logistical tasks and is mostly judgement about priority, access and relationships. This lesson separates the mechanical work from the discretionary work, and explains why the parts that automate cleanly are not the parts the role is valued for.

Generated media makes fabrication cheap in both directions: false material is easy to produce, and real material is easy to dismiss as fake. This lesson covers how verification actually works, why detection tools are the weakest link in it, provenance standards, and the liar's dividend.

Generated prose has a recognisable register, and writers who lean on it converge toward it. This lesson explains why that happens mechanically, what voice is actually made of, how to use a model as an editor rather than a ghostwriter, and why the skill you lose is the one you needed.

Journalism and content writing look like writing jobs and are mostly not. This lesson separates reporting from composition, explains why AI compresses the second and barely touches the first, and covers the verification standard that makes fabricated detail a categorically different problem here.

Supply chain systems handle the normal case; people handle everything else. This lesson covers why exception volume is the real workload, how to triage without automating away judgement, what a control tower does and does not solve, and why visibility is usually a data problem rather than a modelling one.

Moving goods across borders generates a large volume of documents with legal weight. This lesson covers what each one does, why tariff classification and origin are the two decisions that carry real penalty exposure, and how document extraction changes the work without moving the responsibility.

Demand forecasting is the oldest quantitative problem in supply chain and the one most often blamed. This lesson explains where forecast error actually comes from, why the bullwhip effect is a structural rather than statistical problem, and how forecast value added exposes processes that make forecasts worse.

Running these tools inside a design practice: documenting that care happened, the professional indemnity question, who owns generated output, coordination liability across consultants, and the parts of the work a client structurally cannot get from a tool.

Documentation is the biggest consumer of hours in a practice and the clearest place tooling helps. This lesson covers specification drafting and why a spec is a legal instrument, submittal and datasheet review, request-for-information handling, and the coordination problem between drawings and specs.

Architecture and engineering divide into design, documentation, coordination and administration, and AI touches them very unevenly. This lesson maps the split, then explains the professional seal: a licensed person certifying their own judgement, which is why the responsibility cannot move to a tool.

Being small exempts you from very little. This lesson covers the obligations a one-person business still carries: honest advertising after the FTC's AI enforcement sweep, fake reviews, telling customers when they are talking to a bot, data protection, and the judgement about what to keep human when nobody is making you.

Concrete workflows for a one-person operation: handling enquiries without being interrupted, quoting, keeping the back office current, producing marketing that does not sound generic, and knowing when to stop automating. Plus the ladder from template to fully automated, and where to get off it.

A solo owner is not a small version of a company, they are one person switching between every function. This lesson identifies the real constraints, attention and the absence of anyone to check the work, and shows why that changes which AI uses pay off and which quietly create problems.

The professional standards side: Circular 230 diligence and what it demands of a tool user, preparer penalties and where the exposure actually sits, why the tax practitioner privilege is narrower than most people assume, confidentiality of client data, and what remains valuable when research gets fast.

The advisory workflows in detail: framing an issue, running research so the model finds authority instead of inventing it, drafting a memo that does the job a memo exists to do, reviewing documents for tax consequences, and turning a technical position into something a client can act on.

Tax work divides into compliance, advisory, controversy and planning, and AI touches them very differently. This lesson maps the split, then covers the authority hierarchy that governs every tax position, and why the regulation defining substantial authority excludes exactly the kind of output a model produces.

Running AI inside a regulated advisory practice: keeping tools within the firm's capture and supervision system, the confidentiality problem, what you may and may not say about your own AI use after the SEC's first enforcement actions, and the parts of the job a client structurally cannot get elsewhere.

The four workflows that recover the most advisor time: preparing for a review, capturing and documenting the meeting, producing client-ready written output, and digesting research. What each needs as input, where the Marketing Rule engages, and the review that has to happen before anything reaches a client.