On 13 November 2026, e27 published a piece by the head of Stories of Asia (SOA), a small business that runs leadership development through SEED Academy, personal branding through Personal Brand Lab, and professional community building. The author says SOA has no full-time designer and no sales team beyond themselves. Instead, operations staff with no sales training use AI to help draft pitch decks and proposals, and staff without design backgrounds use it to make banners and presentation visuals. The team also uses it for research and first drafts of LinkedIn content.
The claim running through the piece is that AI doesn’t eliminate work. It widens what each person can contribute. I think that claim is mostly right.
Was there a real need? Yes
A small team needs a designer, a salesperson and a researcher, and it can afford none of them. That is a real problem, and it isn’t new. Before AI, the owner either did all three badly or left the work undone.
What changed is that a person with no training can now produce a first draft in each. The article gives no revenue figures or measured productivity gains, so I’d treat it as one team’s account, not a result.
What didn’t change is the checking
The most useful part of the article is where the author admits things went wrong. The team found that some AI-generated training-cost figures and references could not be substantiated. In the author’s words, the problem “wasn’t an obviously incorrect answer. It was an answer that looked ready to use.”
They also found that asking AI to change one element of a design sometimes changed other things they hadn’t asked about. Personal branding drafts needed a person to listen to the client and rewrite the content so it sounded like them.
None of that is a footnote. It is the job. The AI produced something that looked finished, and a person had to decide whether it was.
The skill was always in the directing and the checking
My position on AI is simple. You direct, the AI produces, you check and decide. You are responsible for every output that leaves your hands.
SOA’s story fits that. The operations person who can now draft a pitch deck hasn’t become a salesperson. They have become someone who can brief a draft, read it critically and fix it. The tool didn’t supply that judgement. The author says as much: generating work is getting easier, and exercising good judgement is getting more important.
The order matters here too. Mindset comes first, then skillset, then toolset. A team that expects AI to think for it will ship the answer that looked ready to use. A team that treats AI as a fast drafter, with a named person responsible for the result, will catch the unsupported figure before it reaches a client.
What I’d ask the team
If a business told me it was covering a missing designer, salesperson and researcher with AI, I’d ask two things.
First, which workflow? “We use AI for sales” is a mood. “We draft every proposal from the client’s brief, and the owner rewrites the pricing and the recommendation before it goes out” is a workflow.
Second, who checks the output, and against what? A researcher’s value was never the speed of the first pass. It was knowing which number to doubt. If nobody on the team has that instinct, the AI will give you confident figures and nobody will question them.
One tool, one job
The temptation for a small team is to fill every missing hire with a different tool. I think that is the wrong way round. Each new tool is another interface to learn and another set of outputs to check, and the owner’s attention is already short.
I use Claude for almost everything, and I find one tool used deeply more efficient and more effective than a stack of them. On my own work, Claude does the building, the counting and the checking of the mechanical parts. I make the decisions.
What a business should do
If you run a small shop, clinic or service business and can’t hire the specialists you need, start with one job. Pick the one that costs you the most time or the most missed opportunities, such as proposals, product descriptions or competitor research.
Choose one tool and use it on that job until you know its failure patterns. Write down who checks the output and what they check it against. For numbers, that means a source you can open. For design, it means looking at the whole page after every edit, not only the part you changed. For anything with a client’s name on it, it means a person reads it as the client would.
Add a second job only when the first one runs without surprises.
Source: e27: AI didn’t eliminate our work. It changed what our team could do