Orchestrating intelligence

Prompting isn’t about asking AI a single clever question — it’s about orchestrating intelligence.

Firstly, how intelligent is AI?

VERY. Very, very.

We are at or on the cusp of AGI, Artificial General Intelligence, where AI can understand, learn, and apply knowledge to solve any intellectual task a human can.

Sam Altman, CEO of OpenAi tweeted in 2024 that OpenAI had “achieved AGI internally.”  Microsoft put out a paper in 2024 claiming OpenAI’s GPT-4 AI model exhibited “sparks of AGI,” and Demis Hassabis, the Nobel-laureate cofounder of Google DeepMind, told reporters that the world was “on the cusp” of AGI.

No human can compete with AGI, and we are already there, or very close.

The better the prompt or “conductor’s score”, the better the AI performance. When prompts are vague, AI improvises. When prompts are intentional, AI collaborates.

Secondly, how do you orchestrate intelligence?

The first step in prompting is role and context. Tell the AI who it is and what it’s trying to achieve. For example, “You are a financial planner explaining trade-offs to a cautious retiree” produces a very different outcome from “explain retirement options.” Roles anchor reasoning, tone, and priorities — they’re the difference between noise and music.

Next comes structure and sequencing. Instead of one big prompt, break thinking into stages: ask for assumptions, then scenarios, then risks, then recommendations. This mirrors how humans reason and lets the AI build momentum rather than jump to conclusions. For example, “assume a 3% inflation, growth at 7%, and a 10-year time horizon.”

Finally, prompts as feedback loops. Refine, challenge, and redirect: “What would change if we assumed inflation was 4%?” This is where orchestration really happens. The most powerful AI systems aren’t smarter models — they’re better conversations.

Thirdly, who creates the prompt?

The best prompts will be created by people who understand the industry, its compliance regime, the particular business, and the art of prompting. In a small business, this will probably be senior management. It will be quicker for them to become experts in prompting than hiring an expert “prompter” who still has to learn the business. It also keeps HEADCOUNT down and is easier than learning (say) Excel.

In larger businesses, the process is harder to imagine.  Some prompts will work well across the whole organisation. For example, a more “horizontal” AI agent that summarizes emails from clients and suggests action will work in most departments. A more “vertical” agent that needs human input may only work in that department.

This makes “orchestrating” AI tricky.