A Digiday+ survey of agency professionals just found 73% say their clients don't understand what agentic AI even is — and agencies are pitching it anyway as the reason fees should hold. Here's the proof layer that replaces the buzzword before a client calls it out.
by Ayush Gupta's AI
The problem
Agencies spent the last year building 'agentic AI' into the pitch — a line about autonomous optimization, always-on agents, AI that acts instead of just drafting. The problem is most clients never actually understood what that meant, and agencies kept selling it anyway. A Digiday+ Research survey of agency professionals found 73% say their own clients don't understand what agentic AI is or how it's meant to optimize their campaigns. That gap used to be harmless, because nobody was pushing back. It isn't harmless anymore. Gartner's Jay Wilson put it plainly: the honeymoon around AI and agencies is essentially over — a year ago agencies sold clients on AI platforms promising cost efficiencies and speed, and that story is wearing thin. When a client who never understood the claim starts asking what it actually bought them, 'agentic AI' as a line item has nothing behind it to point to.
The fix
Build an AI system that audits every agentic-AI claim currently in pitch and renewal materials, and replaces each one with a concrete, account-specific proof point the client can verify themselves — instead of continuing to sell the capability as a concept nobody has confirmed they understand.
The Playbook
Pull every 'agentic AI' claim out of current pitch and renewal materials
Before fixing anything, find out how much of the current pitch is actually riding on a term the client can't define. Go through the last two quarters of decks, proposals, and renewal docs and flag every sentence that uses 'agentic,' 'autonomous,' 'AI agents,' or similar language as the justification for value or price.
Here are our last several pitch decks, proposals, and renewal documents:
[PASTE DECK/PROPOSAL TEXT]
Pull out every sentence that uses "agentic AI," "autonomous," "AI agents," or similar language to justify value, differentiation, or pricing.
For each one, tell me:
1. What specific capability is being claimed
2. Whether there's a concrete, account-specific result backing it up anywhere in the same document
3. Whether a non-technical client stakeholder could verify the claim themselves if asked
Flag anything where the honest answer to #3 is no.Replace each unverifiable claim with one real before/after number
For every flagged claim, go find the actual account data behind it. If there's no real number, that claim doesn't survive this process — it either gets cut or gets rebuilt once there's something true to say. The goal is one sentence per claim that a client's finance person could check against their own dashboard.
This is the agentic AI claim we've been making: [PASTE CLAIM]
Here's what actually happened on [CLIENT NAME]'s account that might back it up: [PASTE RAW DATA — turnaround times, campaign changes, output volume, etc.]
Rewrite the claim as one sentence with a specific, checkable before/after number instead of the general capability language. If the data doesn't actually support a claim this specific, say so instead of forcing one.Build a one-page explainer in the language the client's stakeholder actually has
The 73% gap isn't a client intelligence problem — it's a translation problem. Most explanations of agentic AI are written for someone who already works in AI. Build a one-pager per account that explains, in plain operational language, what the AI is actually doing on their account and what stays human — no architecture diagrams, no vendor language.
Write a one-page explainer of what agentic AI actually does on [CLIENT NAME]'s account, for a stakeholder who has no technical AI background and no reason to pretend they do.
Rules:
1. No jargon — no "autonomous agents," no "orchestration," no vendor buzzwords
2. Describe it as: what used to happen manually, what happens now, who still checks it
3. End with one sentence on what never gets automated, full stop
4. Keep it under 300 wordsGive the account team a live answer for 'wait, what does that actually mean'
The moment this gap becomes visible is live, in a call, when someone finally asks the question out loud. The AM needs an answer ready that doesn't sound defensive and doesn't punt to 'I'll have someone follow up.' Rehearse it before it's needed, not after.
Draft a 3-4 sentence answer an account manager can say out loud when a client asks "wait, what does 'agentic AI' actually mean for us?"
The answer should:
1. Translate the capability into one plain example from their own account
2. Not use the word "agentic" in the answer itself
3. Sound confident, not rehearsed or defensive
4. End with an offer to send the one-page explainer, not a deflectionRun a quarterly claim-cleanup pass instead of a one-time fix
New claims creep back in every pitch cycle, usually from whoever built the deck fastest the night before a call. Put a standing check in the proposal review process: no capability claim ships without a proof point attached and a plain-language translation on file.
What changes
Pitch and renewal materials stop leaning on a term nearly three-quarters of clients can't define, and start carrying one verifiable result per claim instead. When a client finally asks what 'agentic AI' bought them, the account team has a plain answer and a number ready, instead of restating the buzzword louder.
Agencies spent the last year building "agentic AI" into the pitch — a line about autonomous optimization, always-on agents, AI that acts instead of just drafting. The problem most agencies haven't confronted yet: a lot of clients never understood what that meant, and the pitch kept going anyway.
A Digiday+ Research survey of agency professionals put a number on it. 73% of respondents said their own clients don't understand what agentic AI is or how it's supposed to optimize their campaigns. That's not a knowledge gap on some abstract industry term — it's a gap sitting directly underneath a claim agencies are using, right now, to justify fees and differentiation.
Why this gap stopped being harmless
For most of 2026, the gap didn't matter, because the story was working on momentum rather than scrutiny. Clients heard "AI agents," assumed it meant something impressive, and didn't push. That's the part that's ending. Gartner VP analyst Jay Wilson described it directly: the honeymoon around AI and agencies is essentially over. A year ago, agencies were selling clients on AI platforms promising cost efficiencies and speed to market. That pitch bought goodwill on faith. Faith has a shelf life, and the same research shows agencies are already feeling it — reduced client budgets and the effects of AI tied as the two biggest industry concerns heading into next year, each cited by 38% of respondents.
Put those two numbers next to each other and the exposure is obvious. Clients are getting more cost-sensitive at the same time they're losing patience with vague AI claims they never understood in the first place. A line in the deck that depends on both client goodwill and client confusion is not a line that survives a harder renewal conversation.
Where agencies get caught
It rarely gets caught in the room where the claim was first pitched — it gets caught a few months later, in a renewal or a budget review, when a different stakeholder who wasn't in the original pitch asks a plain question: "what does 'agentic AI' actually mean for what we're paying you?" The AM who built the original deck is sometimes not even in the room. Whoever is in the room either has a real answer with a number behind it, or repeats the buzzword louder and hopes that reads as confidence. Clients can tell the difference immediately.
The accounts that get hurt worst aren't the ones that never used the term. They're the ones that built real differentiation into the pitch deck and buried it under language nobody asked them to translate — so the actual work goes unverified and uncredited at exactly the moment it needed to hold up.
The fix: trade the buzzword for one number per claim
The fix isn't to stop talking about AI capability. It's to stop asking the client to take the capability claim on faith. Every place "agentic AI" currently justifies value in a deck or renewal doc should get replaced with one specific, checkable before/after result from that account — a turnaround time that changed, a volume of work that shifted, a gap that closed. If there's no real number behind a claim, that's the signal the claim shouldn't be in the deck yet, not a reason to phrase it more impressively.
The second half of the fix is translation, not proof. A one-page account explainer, written for a stakeholder with zero technical AI background, describing plainly what used to happen manually, what happens now, and what still gets a human check — no architecture language, no vendor terms — closes the literacy gap directly instead of hoping the client nods along.
Bottom line
73% of clients not understanding the exact term an agency is using to justify its fee is not a client problem. It's an unfinished translation job sitting inside the pitch. Replace the buzzword with one verifiable number per claim, write the plain-language version before the client has to ask for it, and the next hard renewal question lands on an account team with a real answer — not one repeating "agentic AI" slightly louder.