There’s no AI in team: what to delegate (and what to keep)

High-level takeaways from a talk by Articulate’s CEO and Founder, Matthew Stibbe, on why most AI adoption fails, including five ‘no-regrets’ jobs to hand to AI and five areas you must keep under human control.

If you’ve been around B2B marketing for more than five minutes, you know the pattern:

AI adoption is everywhere.
Outcomes are… patchy.

AI can accelerate a lot of work, but it can’t own outcomes, exercise taste or take responsibility. And if you delegate the wrong things, you simply get a faster mess.

Why most AI adoption projects fail (even when everyone’s ‘using AI’)

The headline numbers look impressive. Marketing teams report widespread AI use, and adoption is rising fast.

But there’s a recurring trend across studies and client work: teams spend a lot of time ‘doing AI’ without generating measurable business outcomes. In practice, people often burn time:

experimenting with tools in unstructured ways
generating content they don’t end up using
redoing work because outputs weren’t accurate, relevant or aligned to strategy

That’s because AI multiplies what it finds. If your process is unclear, your positioning is vague or your workflow is chaotic, adding AI doesn’t fix it; it scales it instead. (We unpack the more pragmatic alternative in 6 tips for AI pragmatists in 2026.)

So the question isn’t ‘Are we using AI?’

It’s ‘Are we using AI in a way that makes us better?’

The five failure modes to avoid

1. The ‘wisher’ prompt

This is the classic: ‘Write me a post about AI trends and make it sound smart.’

The cost of lazy prompting is predictable:

no audience understanding
no strategic context
no guardrails
generic output that could belong to anyone

Better approach: give AI real inputs (interviews, meeting notes, messaging docs, examples, the point of view you want to argue), then use it to shape and draft, not to invent your thinking. Our guide to writing better briefs applies just as well to briefing a model as it does to briefing a writer.

2. Assuming AI always makes things faster

Sometimes AI is fast. Sometimes it isn’t. And that can be fine.

A strong workflow often looks like this:

give the model clear instructions and context
let it run a longer process
come back to review the result

Speed matters less than moving effort away from busywork and toward judgment.

3. The ‘copy-paster’

If you copy-paste AI output straight into a client email, a website or a report, you’re treating a first draft like a final draft.

But the reality is, every AI output is a first draft.

AI doesn’t mind revision. Ask it to rewrite, tighten, remove fluff, and align to your tone. But always review before you put your name on it. If you want a systematic final check, use our 24-step expert proofreading guide.

4. More content, more ‘slop’

When the cost of producing words drops to near-zero, the temptation is to increase volume.

But audiences are getting better at ignoring:

generic ‘AI-shaped’ content
glib answers with no lived experience
confident claims without grounding

If you want a benchmark for what that looks like, compare the difference between ‘SEO content’ and genuine expertise-led thinking. It’s the kind we aim for across our webinar and content programmes. We’ve written before about why AI adds to the noise instead of cutting through it, why fighting AI slop with AI slop is a losing game, and how to find content ideas that demonstrate genuine thought leadership instead.

5. Mistaking confidence for correctness

AI outputs are often persuasive. That’s the danger.

Think of it as a brilliant, fast intern who occasionally lies.

Treat AI output like you’d treat anything going out under your name:

verify facts
cross-check sources
don’t ‘quote by accident’ (i.e. repeating a summary as if it’s a verbatim quote)

Five ‘no-regrets’ jobs to hand over to AI

Matthew shared five areas that are usually safe and high-leverage, if you use them properly:

1. Researching and monitoring

AI is excellent at scanning lots of sources, deduplicating, and surfacing what matters. It’s especially useful when you’re trying to manage information overload on big writing projects.

This is one of the best uses of AI because it expands coverage without pretending to be the decision-maker.

2. First drafts (with context)

AI is great at producing a first draft when you give it real materials:

your point of view
messaging frameworks
audience insight
examples of what ‘good’ looks like
source documents like transcripts and notes

The draft becomes a starting point you shape rather than a finished product you publish. It’s the same principle as the ‘shitty first draft’: less drafting, more polishing.

3. Consolidating feedback

If you’ve got multiple rounds of comments, meeting notes, or stakeholder opinions, AI can:

summarise themes
reconcile contradictions
extract action items
propose a clean next draft

It works best alongside a human process like pair writing, where two people work through a draft together.

4. Audits at scale

AI is tireless for systematic work: SEO checks, QA lists, structured reviews, and pattern spotting. Think SEO website audits, technical SEO audits and the repeatable checks we run in our SEO sprints.

It won’t replace expertise but it can dramatically reduce the manual grind.

5. Mechanical transformation

Turning this into that is a sweet spot:

survey responses → themes
interviews → evidence sets
transcripts → structured articles
messy notes → usable outputs

Five things to keep under human control

Every job can involve AI, but every outcome must be human-owned.

These are the areas where humans have to stay in charge:

1. Judgment

AI cannot take responsibility. You can’t outsource accountability.

Think of a great restaurant: the head chef isn’t cooking every plate, but they are deciding what leaves the kitchen. That’s the role humans play with AI.

2. Data protection

One careless paste of sensitive information can create real risk.

Guardrails matter:

avoid putting confidential client information into public tools
don’t paste API keys
don’t treat ‘chat’ as private by default

3. Accuracy

Fact-checking is not optional. Especially when:

quoting statistics
making claims about markets or competitors
giving advice that will influence decisions

4. Transparency and control

Rogue AI shows up when:

teams adopt tools independently
people use personal accounts instead of company tools
automations run without oversight (and quietly rack up costs)

Making safe experimentation easy with visibility and sensible boundaries.

5. Your human edge

Delegating skills to AI can cause atrophy. It’s one reason expert copywriting still pays for itself, and why a distinctive tone of voice is so hard to fake.

Use AI boldly, but keep the muscles sharp:

keep writing sometimes without AI
keep thinking from first principles
keep practising craft, not just output

The ‘hands on the wheel’ checklist (three questions)

When you’re about to use AI on a task, ask:

Does the AI have the right context?
Is a person owning the outcome (and willing to sign their name)?
Is this use of AI protecting our edge or commoditising it?

In short: keep your hands on the wheel.

Where this lands for B2B marketing teams

Used well, AI helps you do more of the thing that matters:

better thinking
clearer strategy
sharper positioning
stronger content that builds trust

Used badly, AI creates:

more noise
more risk
more generic “content-shaped” outputs

If you want the upside of AI without the downside, start simple: get clear on what “good” looks like, build a repeatable workflow, and keep a human accountable for the result.

That’s exactly how we work at Articulate. We help B2B teams build a Difference Engine® — the strategy and system that makes marketing perform consistently — then use AI to make the engine run faster, not noisier. (If you’re curious, here’s the framework: Difference Engine®.)

If this article resonated, you might also like:

How we approach Brand strategy (so AI has something real to amplify)
What “good” looks like for Signature websites (where clarity beats cleverness every time)
Our view on Expert copywriting (so your POV doesn’t get averaged into slop)
Where we use automation safely, including HubSpot activation

More from our blog on getting AI to work for you:

Want to explore what to delegate, what to keep human, and what an AI-assisted workflow could look like for your team? Book a free, 30-minute strategy session.