Your go-to-market stack just went headless, and the teams best placed to benefit are the ones who have quietly kept their account data in order. On 15 September 2026, two specialist vendors made the same move within hours of each other, and neither of them asked anybody to change CRM.
Demandbase launched Mojo, a B2B marketing agent that defines audiences, writes briefs, and then launches and manages campaigns across Google Ads, LinkedIn Ads, Meta, Marketo, Salesforce and Slack. The same day, Gainsight opened its customer intelligence to Salesforce Agentforce and to Slack through Model Context Protocol connectors, and described the integrations, in its own words, as headless. Two different corners of the revenue stack, one direction of travel: the specialist system stops being a place people log into and becomes context an agent can read and act on wherever the work happens.
That is a genuinely good week for revenue operations. The intelligence that used to sit behind a login your sellers never opened is now available to the tools they already use. The catch, if it is even a catch, is that an agent reading across six systems is only as useful as the definitions underneath them, and those definitions are exactly what a revenue operations team owns. Work that has been hard to fund for a decade just became the most leveraged thing on the roadmap.
What did Demandbase and Gainsight actually ship?
Mojo takes a strategy and turns it into a live campaign, then keeps running it. It picks the audience, drafts the brief, launches across connected channels, and monitors what happens afterwards. When something goes wrong in the plumbing, a broken tracking parameter or an audience built on the wrong filter, it flags the problem and brings a proposed fix back to a human for approval rather than quietly acting on it. Each campaign adds to what it knows about the organisation’s audiences and channels, so the next one starts from a better position than the last. Demandbase chief executive Gabe Rogol positions it as the first of a generation of agents that carry strategy all the way through to execution.
Gainsight’s release points at the other end of the customer lifecycle. Its health scores, product adoption trends, sentiment changes, stakeholder maps and churn or growth indicators now reach Agentforce and Slackbot through MCP, spanning Customer Success, Staircase AI, Skilljar, Customer Communities and Product Experience. Customers can also buy prebuilt agents, including an Expansion Analyst that reads Salesforce opportunities and looks for growth the account team has not spotted. Chief executive Chuck Ganapathi’s framing is that the answer to the retention question is “not another generic agent”, which is a fair description of what specialist data is for.
The commercial case Gainsight puts behind it is worth noting, because it is the kind of number that moves a budget conversation. Across the 2,000 or so companies on its platform, top-quartile SaaS businesses hold median net revenue retention of 113 per cent against 98 per cent in the bottom quartile, and the market values that gap at roughly 24 times revenue against five. Retention intelligence has always been worth something. Putting it where the account team already works is how it gets used.
There is one more detail in the Demandbase announcement that every operations lead will recognise. In a 72 hour sample of its agentic usage, the company logged 1,240 questions from 252 users across 176 companies. List building accounted for around 20 per cent of prompts and campaign return and attribution for 17 per cent. Results were exported to spreadsheets 284 times. Even with an agent in the loop, the answer still left the system to be finished somewhere else. That gap between where the insight lives and where the decision gets made is the one these releases are built to close.
What is revenue operations when the tools stop being destinations?
Revenue operations is the function that owns the systems, data and process connecting marketing, sales and customer success, so that a single view of pipeline and revenue holds across all three. Historically that meant integration work: moving records between platforms, reconciling duplicates, and building reports that stitched the picture back together. In an agentic stack the job changes shape. The integration is increasingly supplied by the vendor through MCP, and what remains is curation: deciding which systems an agent may read, which definitions are authoritative when two sources disagree, and which actions an agent may take without a person approving them. That is a more strategic job than the one it replaces, and it is considerably harder to outsource.
It is also a more visible one. When a marketing agent builds an audience from your account hierarchy and a customer success agent reads your health scores, the quality of those definitions shows up in campaign performance and renewal conversations within weeks rather than in a quarterly report nobody reads. Operations teams have spent years arguing that the foundations matter. The agents make the argument for them.
Why is CRM data quality important when agents read from specialist systems?
CRM data quality matters because an agent acting across several systems has no way to sense that something looks wrong. A human seller who opens an account record with three duplicate parent companies pauses and asks a colleague. An agent building an audience from that same hierarchy simply produces an audience that is a third too small, spends the budget, and reports the result with complete confidence. The cost of an inconsistent account identifier used to be a confusing report. Now it is a campaign that reached the wrong companies, or an expansion signal attached to the wrong opportunity.
The encouraging part is how narrow the required work actually is. Agents of this kind lean on a small number of fields: a consistent account identifier, a parent and child hierarchy that matches how the business sells, a clean opportunity stage model, and whichever two or three fields define your segments. That is a short list, and every item on it improves routing, territory design and forecasting whether or not you ever switch an agent on. This is maintenance with a return attached, not a prerequisite tax.
What is CRM technical debt in an MCP world?
CRM technical debt is the accumulated cost of customisations, integrations and workarounds that made sense when they were built and now make every change slower than it should be. Custom objects nobody populates, three overlapping middleware connections doing one job, automation written around a process that changed two reorganisations ago. Under the old model this debt was mostly a drag on delivery speed. Under an agentic model it also shapes what an agent sees, because an agent reading your org reads the debt along with everything else.
Here the news is unusually good. A standard connector that a vendor maintains is one fewer bespoke integration for your team to own, and the specialist platforms shipping MCP servers are quietly retiring a category of custom work. Microsoft made the same move for Dynamics 365 Sales in July, when seven external data providers became directly available to its sales agents. Salesforce has been extending Headless 360 along the same lines. The practical opportunity for the next quarter is a straightforward audit: list your point-to-point integrations, mark the ones a vendor now covers as standard, and retire them. Most organisations find the list is longer than they expected, and every retirement is a maintenance cost that disappears permanently.
How to align sales and marketing around one set of definitions
Sales and marketing alignment has usually been pursued through meetings and service level agreements. The more durable version is definitional: both functions work from the same account record, the same segment logic and the same idea of what a qualified opportunity is, so there is nothing left to argue about in the first place. Agentic tooling turns that principle into something immediately practical, because a marketing agent and a sales agent drawing on different definitions of the same account will produce visibly contradictory behaviour within days. The fastest way to align two teams is to have them agree the definitions that both of their agents will use.
A workshop that produces one agreed account hierarchy, one segment definition and one list of the fields each side trusts is worth more now than it was a year ago, because the output feeds systems rather than sitting in a document. Run it before you configure anything. Teams that do this first tend to find the configuration itself takes days, because most of the difficulty was never technical.
How to measure CRM ROI on an agentic go-to-market stack
Measuring the return on CRM investment gets easier when agents are involved, because agents produce a clean record of what they did. Start with a baseline before anything is switched on: campaign setup time, the number of manual exports your team performs in a week, time from expansion signal to a conversation with the customer, and the share of accounts with a complete hierarchy. Then measure the same things a quarter later. The comparison is unusually honest, because the agent logs every action it takes rather than relying on anybody’s recollection.
Those 284 spreadsheet exports in 72 hours are a useful example of a metric worth owning. Every export is a moment where the system could not finish the job and a person carried the answer somewhere else. Counting them takes an afternoon, the number is embarrassing enough to be motivating, and watching it fall is a clearer measure of progress than most dashboards offer. Pair it with net revenue retention, and you have a before and after that a finance director will accept.
The Sirocco perspective
We work across Salesforce, HubSpot and Microsoft Dynamics 365, and releases like these two are the ones we most like to see, because they improve a client’s position without requiring anybody to change platform. A headless specialist layer means the choice of CRM matters slightly less than it did, and the quality of the definitions underneath it matters considerably more. That suits organisations who have invested in getting their data model right, and it gives everybody else an unusually clear place to start.
Our advice to clients this quarter is deliberately modest in scope. Pick one specialist system you already pay for, whether that is your marketing platform, your customer success platform or your intent provider, and work out what it would take for a CRM agent to read from it reliably. The answer is usually a handful of fields and one agreed hierarchy. Do that work once and it serves every agent you connect afterwards, on whichever platform you run. That is the rare kind of project where the groundwork compounds instead of expiring.
If you would like a second opinion on where your go-to-market data stands, or on which system to connect first, schedule a consultation and we will walk through it with you.
Get in Touch
If you are working out which go-to-market system should feed your CRM agents first, and what needs tidying before it does, we are glad to talk it through.
