Gil Allouche Says Your Best Marketers Are Doing Software’s Job…And That’s Not Good

Picture a Tuesday at a company with a nine-figure revenue line. Its sharpest marketer, the one who understands the buyer better than anyone, is exporting an audience from one ad platform so she can rebuild it in another. Next comes the weekly performance report, stitched together by hand from four dashboards. If the week goes well, she reaches the campaign idea she pitched a month ago by Thursday afternoon.

Gil Allouche has a blunt view of that scene. “That should embarrass the software industry,” he says, and he counts his own corner of it.

Allouche is the founder and CEO of Metadata, a San Francisco, Calif.-based company that sells AI agents to B2B marketing teams at businesses with $100 million to $10 billion in revenue. His argument is that enterprise growth needs more marketing capacity, and that most companies confuse capacity with more hiring.

The argument arrives as marketing budgets hold flat. Gartner’s 2026 CMO Spend Survey of 401 marketing leaders found budgets flat at 7.8% of company revenue. Fifty-six percent of those CMOs said they lacked the budget to deliver their 2026 strategy, and 54% said they lacked the resources.

Why the tools keep failing

Most marketers already use AI. Fewer use it across the whole job. ActiveCampaign’s 2025 survey of 1,000 marketers found that 82% use AI for marketing, but only 23% use it across all three stages the report tracks, which it calls imagining, activating and validating campaigns. Many marketers use AI for pieces of the job, and the handoffs between those pieces still fall to people.

That is the gap Allouche built the company around, and he has turned it into a rule. “Three people is our design constraint,” he says. “It forces us to build a connected [go-to-market] GTM system instead of assuming a customer will hire someone to bridge every gap.”

The rule differs from some software models that rely on an admin, agency or consultant to help implement the product. By Allouche’s reasoning, designing for three people means the product has to do that bridging itself, or the customer cannot use it.

Ambition gets rationed

Allouche’s diagnosis starts with the hours. Configuring campaigns, rebuilding audiences and assembling reports all take time, and every handoff between people and tools takes more.

“Every handoff claims time that could have gone into a better offer or a useful experiment,” he says. “The team’s ambition gets rationed by its operating capacity.”

Broader research on knowledge work points in a similar direction. Asana’s 2023 Anatomy of Work Global Index found that workers spend 58% of their day on “work about work,” the coordination that surrounds the job instead of the skilled work itself. That figure covers knowledge workers in general rather than marketers alone.

Gartner’s data shows companies paying more for marketing labor: labor’s share of the marketing budget rose from 21.9% in 2025 to 24.5% in 2026. How much of that labor goes to administration in marketing specifically is harder to pin down.

One response is another hire, usually a specialist for each channel: one person for LinkedIn, one for paid search, an analyst to reconcile the numbers. Allouche thinks that instinct is how teams end up bigger without getting faster. “A team should be able to pursue a large market without adding a specialist for every channel,” he says.

What the numbers show

The evidence he leads with is a customer result, not a headcount plan, he says. According to a case study published by Metadata about Zoom, the company’s paid program used less ad spend while producing substantially more opportunities in its customer relationship management system (CRM), with campaigns managed using Metadata’s technology. The same case study says campaign setup across a large number of cities was reduced from weeks to minutes.

Zoom,  ibm, Cisco, Reddit, are the largest customers Metadata cites. The company says it has about 20 published case studies from mid-market companies, the segment where one marketer often runs multiple channels at once.

Metadata also commissioned Forrester Consulting to run a Total Economic Impact (TEI) study, the firm’s standard model for what a technology purchase returns over time. Forrester interviewed Metadata customers and built a composite organization from their data. Over three years, that composite showed a net present value of $3.26 million and a payback period of less than six months. Allouche has said the study also found customers could increase their available ad budget by 18% by cutting the agency spend needed to manage campaigns.

Forrester’s policy is that it does not endorse vendors, clients cannot buy favorable results, and every TEI is clearly marked as commissioned.

“Those outcomes say more about enterprise marketing capacity than a headcount target ever could,” Allouche says. In his view, a headcount target is a guess about how much work a team can do, while a drop in setup time is a measurement of it.

According to Allouche, the system behind those numbers runs on seven AI agents that span strategy, audiences, creative, offers, campaigns, analytics and operations. They are designed to turn buyer signals into target audiences, build landing pages and ads, prepare campaigns, enrich leads and connect performance to pipeline. Metadata says it also exposes 171 tools through Model Context Protocol, which means people and agents can call the same work from AI environments such as ChatGPT, Claude and Codex. In principle, a marketer could request an audience from the assistant already open on her laptop.

The line Allouche says he will not cross is who decides. “People direct the work and approve the decisions that commit brand and budget,” he says.

What the humans do instead

Allouche’s pitch is less about cutting cost than about what a skilled team does with its week back: understanding customers, sharpening the offer, and deciding which markets deserve investment.

“A capable marketer should be able to test a good idea this week without negotiating for another specialist’s calendar,” he says.

Founders may recognize the alternative. An idea gets approved on Monday, then waits for paid search, creative, and tracking before launching, sometimes after the moment that made it interesting has passed. Software was intended to shorten that queue, but Gartner’s 2025 Marketing Technology Survey suggests there is still room for improvement: marketers report using only 49% of the marketing technology capability they already pay for.

The catch

Allouche’s argument carries real risks. The strongest numbers come from Metadata’s side of the table: its own customer stories and a study it paid Forrester to conduct. The company’s roughly 20 mid-market case studies are consistent with the Zoom result, but Metadata published all of them. The Forrester work applies an outside method, though Metadata funded it. Neither is a controlled comparison, and results will depend on how clean a team’s data and process are before the agents arrive.

The wider agentic AI market has a reliability problem, too. Gartner predicted in June 2025 that more than 40% of agentic AI projects will be canceled by the end of 2027 because of escalating costs, unclear business value or inadequate risk controls, and its 2026 CMO survey found that 70% of CMOs say their marketing processes are not mature enough to implement and scale AI. Agents can speed up whatever process they are given, including a bad one.

The approval step, Allouche’s main safeguard, has its own weakness. A reviewer who approves fifty correct changes in a row can stop reading the fifty-first closely. And in large enterprises, the slowest part of a launch is often legal and brand review, which no ad agent removes. A three-person team at a $5 billion company still answers to a compliance department.

There is a cultural cost as well. Many marketing leaders measure their standing by the size of their organization. A model that needs fewer specialists asks them to give up a familiar status symbol in exchange for output they then have to prove.

A bigger market, less babysitting

Allouche does not frame any of this as doing more with less. His version keeps the team the same size and hands it a larger market.

“Give your best marketers a bigger market to pursue,” he says. “Stop giving them more software to babysit.”

Entrepreneur Media was not involved in the creation of this content.