Agentic CRM is coming. Ten quotes that show where it's heading.

Agentic AI inside CRM platforms is coming, whether we are ready for it or not. The landscape has shifted from 2025 into 2026 as more powerful AI models were released and more people cottoned on to just what AI can do.

Vendors have shifted their marketing from "AI-assisted" to “autonomous” or “agentic”. Analysts who review the market have moved from "AI is a feature" to "AI is the architecture." And critical voices (including myself - a year ago I would have told you AI can’t code as well as a human) have moved from "this won't work" to “this is probably the future and we need to sort out our data architecture pronto!”

But don’t just take my word for it. Here are ten quotes from various sources, including analysts, practitioners and vendors who are betting on this agentic future.

They paint a clear picture of where CRM is heading, and what the current bottlenecks are.

1. The execution layer is moving

"The execution layer in software is moving from humans to agents."

Andrew Bialecki, Klaviyo co-founder and co-CEO, in Klaviyo's investor announcement (March 2026).

Klaviyo positioning themselves as "the autonomous B2C CRM" is the loudest signal yet. When a CRM that sells the ability to build workflows now bets on agents instead, the conversation isn't whether agentic CRM is coming, it's who builds it well/first.

2. Agents that work out their own steps

"agents themselves use reasoning and planning to dynamically decide how to execute work."

Leslie Joseph, Principal Analyst at Forrester, in Forrester's 2026 automation predictions (November 2025).

No longer will agents just be there to write copy, generate designs, or build one-off triggered journeys with the marketer guiding. Instead, the agents will be behind the scenes pre-empting what is needed before the marketer even knows.

Most of what's shipping in CRM today is still the former dressed up as the latter for marketing purposes. The vendors that genuinely decide how to execute, rather than which suggestion to show, are still rare.

3. The end of channel-based marketing

"This marks the end of channel-based marketing as we know it."

Emily Weiss, Senior Principal Researcher in Gartner's marketing practice, on Gartner's prediction that 60% of brands will use agentic AI for one-to-one interactions by 2028 (January 2026).

Gartner reckons the unit of marketing stops being the channel and the campaign and becomes the individual journey an agent runs end to end. This is a version of ‘omni-channel marketing’ we’ve been pitching for years, but agentic makes it all the more important.

You have to understand where a user is in their shopping journey and have an agent take action on their unique situation, rather than slotting them into a pre-existing, generic triggered campaign.

This also influences how you set up your marketing teams. Rather than domain-specific teams (acquisition, loyalty, CRM, CRO), cross-functional marketers who understand all of these in-depth and help drive the agents, I believe, will become highly valued.

4. The consumer side is going agentic too

"Agentic AI will change the top of the shopping journey first. The moment someone decides they need something, they'll no longer open ten tabs or read 20 reviews."

Kaare Wesnaes, head of innovation at Ogilvy North America, quoted in eMarketer (January 2026).

This one I’m not sure about. Personally, I don’t trust agents enough. I want to browse all my options, research in-depth and then pick what I want. Will I trust an agent to know me well enough to do this? Not yet. But I can see it working for some purchases.

Regardless, the chat agents still need an underlying decision agent to call upon to understand who the user is and what their relationship with your brand is. They can say “help me find a new pair of trainers” and your agent can call a decisioning agent to understand that they prefer Reebok over Nike, typically buy in the £100 range, have looked at running tops recently, and already have two pairs in their basket they abandoned a week ago. This knowledge should influence what the chat agent returns.

And I think the channel (chat) should remain separate from the logic (decision agent). That way, any channel can get the same decision.

5. AI doesn't stall because models are bad

"AI doesn't stall because models are 'not good enough.' It stalls because data architecture lags ambition."

Tobie Morgan Hitchcock, InfoWorld (December 2025).

This is a big one. Agents need data in a structure that works for THEM, whereas many data warehouses or app databases are built in a way that works for app requests or data scientists where speed doesn’t matter.

A lot of work needs to go into reconstructing, aggregating, and generally deciding on the optimal shape of data for agents to use.

6. The data drought has accelerated

"None of it changes without data. And we're really getting to that realization that the data drought we've been warning everybody about has actually accelerated because AI is a voracious eater of data."

Liz Miller, VP and Principal Analyst at Constellation Research, in CX Today (December 2025).

AI hasn't fixed the data problem, it has exposed it. Tools that worked on noisy, partial data when a human was in the loop fail loudly when an agent is making decisions per-customer per-session.

7. Bolting AI onto old systems doesn't deliver

"While AI is boosting individual productivity by over 10%, these gains often evaporate because the AI is being bolted onto outdated, people-centric systems."

Uvance Wayfinders, CIO (March 2026).

This is partly due to the point above about data not being architected in the correct way, and partly because existing vendors aren’t willing to tear down their existing way of doing things (which is generating all their ARR) and completely re-architect their systems. That will be costly.

Unfortunately if they don’t I think their lunch is going to be eaten by smaller players. And even if they acquire (looking at you Insider/Bluecore) figuring out how to integrate two big systems is a multi-year job.

Hence why I think Hyp has a shot ;).

8. Fragmented data makes confidently wrong decisions

"An agentic AI tool operating on fragmented, platform-reported, non-identity-resolved data will make fast, confident, wrong decisions."

Sushil Goel, Layer Five (April 2026).

More evidence that you NEED to get your data layer in order to maximise how effective agents can be.

9. CDPs that refresh overnight aren't fit for agents

"Agents can only act on what they can see. Most customer data platform (CDP) pipelines refresh overnight, which means the agent is making decisions on stale data."

Chris Baldwin, Insider (April 2026).

The batch-refresh model that powered the last decade of personalisation doesn't work with agentic workloads. Whether the established CDP vendors can rebuild themselves for sub-minute freshness remains to be seen.

Some pieces of data, such as number of purchases, don’t change minute to minute. The trick is that when it DOES change, the agent is informed ASAP.

10. Who owns the consequences?

"If AI agents are increasingly making marketing decisions, who ultimately owns the consequences when those systems fail, bias outcomes, optimize toward the wrong goals, or simply produce mediocre work at scale?"

Ronan Shields, Digiday (May 2026).

Whilst selling AI systems for the past 5 years the question of observability and control would always come up. In the next few months I believe people will be willing to concede some control and observability but only if they are seeing tangible results.

And yes, the agents will make mistakes. Just like a human can make mistakes. To minimise those mistakes the marketers will need to be on top of things and keep an eye on what the agents are doing (difficult at scale). A good agent harness will be key.

Read together, these quotes say something pretty specific. Agentic CRM is the destination, good data is needed to get there, and marketing teams will need to relinquish some control and change how they work. And the platforms with the loudest agentic announcements aren't necessarily the ones with the cleanest substrate underneath.

To win in the coming years you need to:

• get your data sorted so agents can use it effectively
• consider cross-functional marketers who are highly skilled at using AI
• be willing to concede some control for results
• be willing to experiment with your processes continuously, and AI is continuously changing and no-one knows best practices yet.

Good luck!