Intent Data Just Moved Out of the Dashboard and Into the Workflow. Now Someone Has to Govern It.

On August 20, Intentsify and Clay announced an integration that makes Intentsify's Buyer Intelligence available through Clay's data marketplace and Signals product.^[1]

That is an integration note. It reads like plumbing. It is also the clearest example yet of a shift that changes who is accountable for GTM decisions at a lean B2B company, and most teams will not notice until something automated goes wrong.

What actually changed

Intent data has historically lived in its own portal. Someone logged in, looked at a surge report, exported a list, normalised it, and then the team argued about whether the data was any good. That argument was slow, and its slowness was a safety feature.

Now intent lands in the same tables a team already uses for enrichment and outreach sequencing. Account-level intent and persona-level signals sit next to everything else, available to the same automations and agents. The announced workflow patterns are ordinary and that is exactly the point: monitor a target-account list with weekly reporting on topic spikes, check intent on demand during enrichment, discover net-new accounts surging on relevant topics, identify the professionals tied to those researched topics. (Business Wire via Morningstar, August 2026)

What changed is not access to intent. It is the latency between a signal appearing and real work happening because of it, and therefore who owns the consequences.

Intent vendors are pursuing this deliberately: be everywhere teams already work, rather than winning by pulling users into a dedicated interface. For a team already running an orchestration layer, the incremental cost of trying intent data drops sharply, and the incremental cost of getting it wrong rises just as sharply. (MarketScale, August 2026)

The debate moves from "is this data good" to "what do we do when it changes." That is a much harder question, and it belongs to whoever owns the workflow, not to the vendor.

Pressure-test the scale numbers

Intentsify says its dataset is built from 1.1 trillion monthly intent signals across nine source types, covering 4.2 million accounts currently in market and more than 33,000 topics. Clay's own framing is that it unifies customer and market data and connects to more than 200 third-party data vendors.^[2]

Those are the company's figures, and they are worth reading as a benchmark for the category rather than as evidence of fit. Scale at that magnitude tells you almost nothing about whether the data will work for a mid-sized B2B company selling into three verticals in the DACH region.

The questions that do tell you something:

• Which regions carry enough signal density to route on? A global topic count is meaningless if your addressable market sits in five European countries and the coverage there is thin.
• Which industries resolve reliably to the accounts on your list? Ask for a match test against your actual target-account file, not a sample they choose.
• Which buying-group roles appear often enough to act on? Persona-level signal that surfaces one job family and nothing else is a report, not a routing input.
• What is the refresh cadence, and what does "currently in market" mean in days? In-market is a claim with a decay curve behind it, and the curve is the part that matters for outbound timing.

Map the coverage to your ICP before you map it to a workflow. A vendor citing coverage at that scale has made the diligence conversation easier, not unnecessary.

Account-level intent flattens a buying committee

There is a structural reason persona-level signal is the part worth paying for.

Complex B2B purchases now involve roughly 11 stakeholders on an average enterprise deal per Gartner's B2B buying research, with 11 to 14 internal stakeholders on the most complex ones, each researching independently and often anonymously long before anyone raises a hand.

Account-level intent compresses that entire committee into a single row that says the account is warm. It cannot tell you that security is researching your compliance posture while finance is researching your competitor's pricing, which are two different plays with two different owners and two different pieces of content.

It gets more awkward when the same account spikes on competing topics in the same week. Under account-level routing, that account gets assigned once, to whoever the rule names first. Under persona-level routing, it can legitimately be two motions running in parallel, which is closer to what is actually happening inside the buyer.

Decide in advance which of those two behaviours your system produces. Otherwise you will find out from a rep who got an account they had no context for.

The governance runbook to write before you automate

Four artefacts. None of them takes longer than an afternoon, and all four are cheaper to write now than to reconstruct after a bad quarter.

1. A written, versioned definition of an intent spike. Include default thresholds by topic, by persona and by region, and put a date and an owner on the file. "Spiking" is about to become a term that triggers spend, and undefined terms that trigger spend are how budgets disappear quietly.

2. An explicit list of which workflows may act automatically. Routing, scoring, and outbound messaging are three very different levels of exposure. Write down which ones fire without a human, and which require approval. The default in most orchestration tools is that anything you can build, you can schedule.

3. A reconciliation rule so sellers see the reason inside the CRM. If a rep has to open a second tool to learn why an account was prioritised, they will not open it. The reason has to travel with the record.

4. Metering and budget rules for on-demand intent checks. This is the one teams miss. When an agent calls enrichment at scale, per-check costs compound in a way that a monthly platform fee does not prepare you for. Set a cap, and alert on the cap rather than discovering it on an invoice.

The one-sentence test

A signal that cannot be explained to a seller in one sentence is a signal that will be ignored.

That is the durable filter, and it survives every vendor change. "This account had three people from IT security reading comparison content about our category twice last week" is a sentence a rep can open a call with. "Composite intent score 82" is not, and reps correctly treat it as noise.

Run every proposed automation through that test before you ship it. If nobody on the team can produce the sentence, the workflow is not ready, whatever the data quality is.

What to bring into quarterly planning

Two decisions, and they are budget decisions as much as technical ones.

First, whether intent becomes a production data source with a named owner or stays an experiment someone checks occasionally. Production means definitions, thresholds, an audit trail and a person accountable for the rules. Experiment means none of that, which is fine, as long as nothing automated fires off it.

Second, where it sits in a shifting budget. Industry projections put ABM's share of marketing budget growing from roughly 13% to 18-21% as programmes mature past the experimental phase, with AI tooling and orchestration at roughly 14-17% of marketing budget by the end of 2027. Those are projections rather than measured facts, but the direction is consistent: more of the budget is moving into the layer where these decisions get automated, and comparatively little of it is earmarked for governing them.

The teams that come out of this well will not be the ones with the most signals. They will be the ones who can explain, in one sentence, why the system did what it did.

That is the standard Nukipa is built to: a GTM system where every automated decision carries a reason a human can read. If you want to see what that looks like against your own account data, test Nukipa.