Looking both ways.
And not being run over by AI.

For most organisations, being AI-first has stopped being a choice and become a direction. This is about the second half of being AI-first, the half those programmes rarely reach, and why committing to AI without it leaves an organisation exposed on ground it cannot see.

The logic holds across sectors. Buyers are moving to AI, efficiency is being found there, and the next phase of growth is being designed around it. The programmes are underway, the budgets committed, the internal capability building. None of that is in question here.

01The AI-first case, and the half it misses.

Few make the AI-first case as directly as Andrew Dimitriou, Global Chief Client and Growth Officer at DEPT. If you are not an AI-first agency, you will fail, he says, and he argues from a record: DEPT reports 230% revenue growth over 3 years, with AI cited as a major driver, and its top 25 clients, about half the business, growing close to 30% year on year.

"They do not need AI, they need growth, and AI is the route to it."

02Looking outside-in.

Consider the other direction, the one that has received limited consideration, a direction few boards have turned to face. To be AI-first is to be represented by AI. The same systems an organisation is racing to deploy internally are, at the same moment, reading all it has ever published and assembling the account the market acts on. That account is drawn from the full footprint built over years, much of it forgotten. It is not drawn from the homepage the board signed off last quarter.

An organisation is now read by AI the way it is read by customers, partners and regulators. What AI reads, and whether the organisation can stand behind it, decides how ready it actually is. Moving AI-first while looking one way only risks being run over by the thing the organisation set out to lead.

AI readiness runs on 2 axes, and they are not the same work. The first is internal maturity, the agents and tooling and productivity Dimitriou describes, the axis most boards have funded and can report against. The second is external readiness, how AI understands and represents the organisation to the market, and it has received limited consideration. An organisation can be fully AI-first inside its own walls and still be described by the AI its market relies on in terms that are inaccurate, incomplete or unsafe, because internal capability does not govern external representation.

03The market has already moved.

Buyers have already moved, whether or not the organisation has. $262bn of 2025 holiday spend was accounted for by AI and agents (Salesforce, 2025 holiday shopping data). 58% of consumers now use generative AI for product and service recommendations, up from 25% in 2023 (Harvard Business Review, 2025, survey of 12,000 consumers).

The intermediary reading the organisation is increasingly the one the buyer hears from, and 85% of brand mentions in AI search come from third-party content. The account of the organisation is being written largely by sources beyond its direct control, layered on top of the material it published itself.

04What the intermediary reports carries cost.

57% of organisations cite AI errors, misinformation and hallucinations as the leading AI threat (Gallagher, 2026 AI Adoption and Risk Benchmarking), and 1 in 5 insurance respondents had experienced loss from AI-related risks in the past year (Gallagher client survey, Q1 2026). For those respondents the cost has already been counted.

05The larger share sits closer to home.

It is tempting to place the exposure in what other people publish about the organisation. The larger share sits closer to home.

AI reaches outdated material and returns it as current, held inside the estate the organisation put online over the years and gradually lost sight of.

06An estate larger than the team can see.

That estate is larger than most digital teams can see. 41% of it is undocumented, unknown to the organisation's own digital team (P&C / Sitemorse risk profiling, 2017 to 2023, covering more than 119 million websites). The teams closest to it already recognise the scale of the problem.

It keeps forming on its own: campaign sites left online after the campaign ends, withdrawn products whose PDFs stay published, partner and supplier copies of content, regional and department sites built outside the digital team. Web teams have raised this for years, and what AI changed is to move the problem from a housekeeping matter to the boardroom.

07When forgotten content becomes the position.

This is not a forecast. In 2024, a passenger who had just lost a family member asked Air Canada's chatbot about bereavement fares and was given the wrong answer. A tribunal held the airline accountable for what its own system had told him, and it paid. The content nobody was watching had become the position the organisation was forced to answer for.

Dimitriou makes a point that reaches further than he draws it. Nearly 95% of AI pilots fail to reach scaling, he notes, because the measurement frameworks were never set up, and he cites a McKinsey finding that only a small share of companies see material EBITDA impact from AI, which he puts down to weak business-case design rather than the technology. External representation is a dimension few organisations measure at all. Money spent amplifying that surface before it is confirmed accelerates the circulation of misinformation rather than correcting it.

08Digital Debt, and the order of the work.

The order of the work follows from this. The landscape comes first, and with it the scale of the Digital Debt inside it, graded by value and risk so the forgotten parts can be found, then remediated or removed. After that the position has to be held, because the footprint does not stay still, and neither does the account AI builds from it.

5 questions leadership is challenged to answer about its own online presence

1. What is out there?
2. Who owns it?
3. Is it necessary?
4. Is it current and accurate?
5. Who is managing it from here?

Once the landscape is understood, those answers are grounded rather than assumed. This is the assessment AAAnow has carried out from the outside in for 25 years, and the research cited here is AAAnow's own. AI readiness, in this sense, is the measure of whether an organisation can stand behind what AI reads about it, not of how advanced its internal tooling has become.

09Where this leaves the platform.

The estate AI reads is the estate a content platform is built to hold. There is a fuller argument here, and it deserves a piece of its own: AI has turned digital sprawl into a board-level exposure, which is the strongest case yet for consolidating a scattered estate onto a single platform. More on that to come. For now the immediate point for platform vendors, and the agencies that build on them, is narrower. The partner that can show a client what AI reads holds a conversation no rival is having, and the platform behind that partner is the one the work stays on.

The instinct is right. So is looking both ways.

Being AI-first is the right instinct. It is also the moment exposure runs highest, because it commits the organisation to being represented by systems reading a footprint it has not looked at. An organisation's Digital Debt is measured by what AI reads across all it has built online over the years, much of it forgotten, not by what it believes it presents today.

Looking one way gets an organisation across the road on a quiet morning. Looking both ways is what keeps it crossing as the traffic builds. Control is the foundation of AI readiness.

Sources.

1. BW Marketing World. If you're not AI-first you will fail, says DEPT's Andrew Dimitriou. Supports the AI-first argument, the 230% 3-year revenue growth, the top-25 client growth, and the roughly 95% of AI pilots that fail to scale. bwmarketingworld.com/article/if-you-re-not-ai-first-you-will-fail-says-dept-s-andrew-dimitriou-611899 (opens in a new tab)
2. Salesforce. 2025 holiday shopping data. Supports the $262bn of holiday spend accounted for by AI and agents. salesforce.com/news (opens in a new tab)
3. Harvard Business Review, 2025. Survey of 12,000 consumers. Supports 58% of consumers using generative AI for recommendations, up from 25% in 2023. hbr.org (opens in a new tab)
4. Gallagher. 2026 AI Adoption and Risk Benchmarking, and Q1 2026 client survey. Supports the 57% leading-threat figure and the 1 in 5 insurance respondents reporting AI-related loss. ajg.com/news-and-insights/features/ai-adoption-and-risk-benchmarking-2026 (opens in a new tab)
5. AAAnow Research, December 2023 to January 2026. P&C (human misinformation), 2014 to 2023. Supports the 79% root-cause figure. aaanow.ai (opens in a new tab)
6. P&C / Sitemorse risk profiling, 2017 to 2023. Covering more than 119 million websites. Supports the 41% unknown-estate figure. sitemorse.com (opens in a new tab)
7. Sitemorse survey, 2021, 2023, 2024, 2025. Supports the 92% of web managers recognising the unmapped landscape and uncontrolled PDFs as a challenge. sitemorse.com (opens in a new tab)
8. British Columbia Civil Resolution Tribunal. Moffatt v. Air Canada, 2024. Supports the account of an airline held accountable for its chatbot's answer on bereavement fares. canlii.org (opens in a new tab)