Council Post: The 800-Millisecond Question Your Marketing Stack Can't Answer

Neerav Vyas is Vice President at Capgemini for Customer First: Data & AI and a two-time finalist for the AIconics AI Innovator of the year.

getty

Somewhere in your funnel right now, a customer is asking your brand a question. She isn't filling out a form or clicking a banner. Her AI assistant is doing the shopping: researching, comparing and deciding on her behalf in well under a second. Your marketing stack, meanwhile, syncs on a nightly batch. It will formulate its answer in roughly 24 hours. She will be someone else's customer by breakfast.

That gap, milliseconds versus overnight, is the defining architecture problem in marketing today. And it explains why so many enterprise marketers are rebuilding their stacks around a single word: composability.

Two forces broke the old model.

One changed how marketers collect data, while the other changed how fast they have to use it.

1. Privacy

Third-party signal loss has made marketing structurally more expensive: In our client work, I routinely see cost-per-outcome rise sharply while conversion rates fall, across nearly every channel and industry. The response, correctly, has been a rush to first-party data. But first-party data at enterprise scale exposed a flaw in how the legacy marketing clouds are priced. Most charge by profile volume or consumption. Feed them everything, and the meter spins.

The result is profile bloat: tens of millions of duplicate, stale or unconsented records sitting inside a platform that bills you for each one. Marketers are paying overage penalties for their own poor data hygiene at rates the vendor set. Strip away the invoice language and the arrangement is stark: You are paying someone to store your own customer data and paying more every time you make new customers.

2. Agentic Experiences

This is the one that changes the physics. Unlike a chatbot reading a script, an AI agent reasons, plans and acts across systems. It answers a service question, checks inventory, applies a loyalty offer and books the follow-up in one conversation.

That agent is only as good as the context it can retrieve in real time. An agent grounded in a profile copied into a proprietary silo 18 hours ago will confidently do the wrong thing: recommend the sofa she returned, upsell the plan they just canceled. The customer never learns the word for your architecture. She doesn't have to. She feels it instantly, long before she can explain why. In an agentic world, data latency is no longer a back-office nuisance. It is the customer experience.

It cuts both ways. The same architecture that grounds a customer-facing agent also powers the internal ones: agents that build audiences, run experiments, generate creative variants and hand governed briefs to agency partners without a ticket queue. The productivity story and the experience story are the same story. Both depend on data the enterprise actually controls, modeled for machine consumption and retrievable in milliseconds.​

Composability is the answer, but it is also a spectrum.

Composability, in plain terms, means the data stays in your cloud data warehouse or lakehouse, the system of record you own, while best-of-breed tools connect to it rather than copying from it. The tools become what they should have been all along: Rent them, use them, swap them, and the customer relationship never moves an inch. Zero-copy access and reverse ETL replace the old ritual of duplicating customer profiles into every application that wants one. You pay for intelligence and activation, not for storing data you already own three more times.

But here is where the discourse goes wrong: Composability is not a binary conversion, and it is not a religion. It is a spectrum, and every position on it carries trade-offs.

At the fully packaged end, the vendor is effectively your infrastructure team. You get speed to launch, no-code workflows and mature orchestration. You also inherit their pricing model, their roadmap and their copy of your data. At the fully composable end, you get architectural freedom, warehouse-native economics and agent-ready latency. You also sign up for real engineering headcount, infrastructure you now manage and integration work the suite used to do for you. Lower license fees can quietly become higher people costs if you choose a position your organization cannot staff.

The honest question is not "suite or stack." It is: Where on the spectrum does each workload belong, given your data maturity, your engineering capacity and the clock speed your use cases demand?

The suites still matter.

None of this is an obituary for the major marketing clouds. Their orchestration, journey and channel capabilities remain genuinely hard to replicate, and the vendors themselves are embracing zero-copy connectivity because their best customers demanded it. The enterprises getting this right are not ripping out their suites. They are repositioning them. The warehouse becomes the single ledger of customer truth; identity resolution and data hygiene happen there, and only clean, consented, activation-ready profiles flow into the suite.

That one architectural move can help do three things at once. It can collapse the profile counts driving your overages, often worth millions annually. It can make the suite better at its actual job because it finally runs on trustworthy data. And it can build the real-time context layer that every agentic experience, customer-facing or internal, will likely draw on for the next decade.

The last era of marketing technology asked one question: Which platform should we buy? The agentic era asks a harder one: Which capabilities do we own, which do we rent and can our data move as fast as our customer now does?

Answer that deliberately, workload by workload, and the suite becomes a tool again instead of a landlord. Because none of this was ever really about technology. It is about the oldest job in marketing: showing up in the moment your customer looks your way. That moment now lasts 800 milliseconds, and it belongs to whoever can be in it.​​

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?