Track record across the work
Most organizations now have AI tools, but far fewer have AI their people actually rely on. These are the problems I hear about most often. I have worked through each of them, and the systems I built in response are still in daily use.
Demos impress, then stall. I start from the real workflow, so the system is part of the job from the first day.
Generic answers and made-up facts come from missing context. I give every agent one approved picture of the company.
Years of know-how sit in folders nobody can search. I turn scattered files into records that people and AI can question directly.
Pursuits live in personal spreadsheets, so nobody can say what is active, what it is worth, or where it stands. I bring them into one register with a lifecycle and a scorecard.
Adoption is a skills problem, and skills can be taught. I show people how to direct AI rather than just ask it questions, and I stay with them while they learn.
Control works best when it is built into the workflow from the start. Every system I build has permissions, review steps and a named person who signs off.
A voice can be written down and held to. I give AI the house style, examples and limits, and people keep the final say.
Most of a bid is setup, reuse and assembly. AI can carry that load, so experts spend their time on the parts only they can write.
Every post is a one-off effort, so output stalls and quality wanders. I build the register and the workflow once, so production runs on a system, not on heroics.
AI video can meet a professional standard when someone directs it with care, and I produce consistent, on-brand series in-house.
See how each one was built: AI systems →
Why this matters: 95% of organisations report no business return from generative AI (MIT NANDA, 2025), and over 40% of agentic AI projects are predicted to be cancelled by the end of 2027 (Gartner, 2025). Every system on this site was built with those failure points in mind.
Each of these began with a real bottleneck that was slowing people down, and a person reviews every output before it is used.
Turns a new RFP into a folder structure, branded templates and an 80 to 90% complete first draft in approximately 30 minutes. Setup used to take 1 to 2 days.
View system → Built with ClaudeDrafts company social posts and tracks each one through review, approval and publishing, so every post has one tracked record instead of a scattered email thread.
View system → Microsoft Copilot StudioAgents for bids, marketing, Google Ads, cybersecurity and sales outreach. The bid agent drafted a 35-page RFP response live, in a single demonstration.
View system → Weekday digestNew public-sector opportunities from 18 public sources, delivered every weekday to the people who make go/no-go calls.
View system →Most AI work stops at the prompt. The systems on this site draw on something deeper: one structured knowledge base that gives every agent the same accurate picture of the company.
I built the company’s AI knowledge base, drawing on knowledge from people across the business, and wrote its agent identity and tone-of-voice standard. Its 86 interlinked Markdown files are the shared brain behind the company’s AI agents and Claude projects. A routing layer sends each question to the files that hold the answer.
Each project and agent connects with defined permissions. Some can only read; others can write approved updates back. As more departments use it, the knowledge base grows, and every connected system gains from what one team adds.
Why it matters. Anthropic’s engineering team calls this discipline context engineering: deciding what information a model works from, because “context must be treated as a finite resource.” A curated, shared knowledge base is how that discipline scales across a whole company.
One knowledge base, many connected systems. Each connection has its own read or write permission, and new projects plug into the same foundation.
Most people use AI to answer questions. I work with it the way a director works with a crew: set the vision, give clear direction, push back when the first answer falls short and take responsibility for the result. Every system on this site was built that way, and it is a skill any team can learn. This 54-second film I wrote, directed and edited shows the idea.
Decide the result, the audience and the standard before anything is made.
Give AI the facts, examples and limits you would give a capable new team member.
When the answer is "that can't be done", ask for another route. There usually is one. When the limit is real, listen.
Nothing ships until a person has reviewed it. The director owns the result.
A free course and eight short guides I wrote, from directing AI well to five problems teams are working on now. Start with the full course, or download the guide that fits the problem in front of you today.
A free, self-paced course built on these eight guides, in order. 8 modules, field exercises, knowledge checks and a certificate of completion. No sign-up.
Each guide is two pages, free to download and free to share with your team. Guides 1 to 3 teach the method. Guides 4 to 8 each take on a problem teams are facing right now.
Guide 1 · The methodTake the director's chairFive habits of a good director, a brief template with a worked example, what to do when AI says "that can't be done" and a checklist before you share anything.Download PDF
Guide 2 · The mindsetThink like the directorWhy people stay in their crew role with AI, six mental blocks such as the ELIZA effect and functional fixedness, and a 12-point checklist to move into the director's chair.Download PDF
Guide 3 · The wide viewSee the whole setKnow a little about every department and stay curious: a 15-minute way to learn a new field, five curiosity habits and when to bring in an expert.Download PDF
Guide 4 · The adoption gapGet the crew workingFor teams with AI licences that few people use: five reasons adoption stalls, a 30-day plan for one team and the habits that keep it going.Download PDF
Guide 5 · The context gapGive the crew the scriptFor agents that give confident wrong answers: the five kinds of company knowledge an agent needs and how to build a knowledge base in two weeks.Download PDF
Guide 6 · Brand voice at scaleKeep one voiceFor AI content that all sounds the same: a one-page voice file, five checks before anything publishes and an approval path that holds at volume.Download PDF
Guide 7 · Proposal bandwidthMore bids, same crewFor proposal teams facing more RFPs with the same people: where bid time goes, an AI-ready bid library and where people must stay in charge.Download PDF
Guide 8 · Series continuityKeep the series on modelFor AI video that drifts from clip to clip: what to lock before you generate, a scene-by-scene plan and five checks for every clip.Download PDF
The two-minute video behind the first guide: why good prompts still produce bad results, and the shift that fixes it.
The ELIZA Effect: Why Good AI Prompts Still Produce Bad Results
I wrote and produced this two-minute video to change how people think about AI. Weak results rarely come from weak prompts. More often they come from the ELIZA effect, the old human habit of treating a machine as if it already understands us. The video names that habit and the shift that corrects it, which is where my whole approach begins: stop asking AI questions and start directing it.
Read: Take the director's chair →
Produced for MYRA Systems Corp. The video is MYRA’s property and plays from the company’s official YouTube channel.
If a guide helps you or your team, I would love to hear how it goes.
The record behind the work, beyond the day job.
The tools I reach for most. There are more behind the work; see the About page for a wider list.
Product names and marks belong to their owners and are shown only to list tools I use. No endorsement or partnership is implied.
Practical articles on applied AI, content operations and video production. All articles
My creative work covers writing, design, illustration and animation. It includes DJC Media, the independent studio I run, whose channel has passed a quarter of a billion views. The studio follows the same directing approach as the systems above.
