Gartner estimates that over 40% of agentic AI projects will be canceled by the end of next year due to escalating costs, unclear ROI and a lack of governance over risk. That stark forecast has not stopped platform vendors from racing to fill marketing org charts with non-human hires — and it should not stop marketers from planning for them, argues a new analysis from MarTech.

Optimizely announced Virtual Teammates at its 2026 Opticon conference earlier this month. The interface presents a directory marketers can hire from, similar to reviewing a contractor's resume on a job site. Available roles include a chief of staff, an SEO and AI search analyst, a marketing analyst, a personalization strategist and a CRO manager. Each virtual marketer has its own identity and permissions, can run on a schedule or a trigger, maintains context across projects and includes an audit trail linked to that agent.

Optimizely is not alone. Earlier in the year, Treasure AI released a Marketing Super Agent it describes as operating like a marketing department, orchestrating other agents inside a governed workspace. HubSpot shipped more than 20 Breeze agents and called them digital teammates, taking a shot at AI tools that behave like expensive interns. Asana's AI Teammates also come in role shapes, including a Campaign Strategist, with the team holding permissions and data access.

All four vendors have converged on the same architecture: identity, permissions and an audit trail. According to MarTech, when vendors converge like this, it means a category is forming — and virtual teammates are likely here to stay in 2027 and beyond.

The human half of the hybrid workforce

EY recently announced a $100 million rewards program for employees who demonstrate critical thinking skills in their work alongside AI tools. Individual spot awards top out at $500, while team awards range from $10,000 to $25,000. Qualifying behaviors include judgment, business acumen, adaptability, collaboration and experimentation with the technology itself. EY, in effect, valued team play at 20 to 50 times the individual prize.

EY is not an outlier. KPMG is reimagining its audit intern program with an emphasis on critical thinking, and PwC is teaching its people AI proficiency alongside empathy and innovation. Within 12 months, three prominent consultancies have redefined what they reward and clarified expectations for humans in hybrid human-AI environments. MarTech advises companies pursuing similar models to assign clear accountability, with identifiable points of responsibility for the stewardship of AI-generated output.

Agentic reality check

The same Gartner report estimates that of the thousands of vendors selling agentic AI, roughly 130 are actually building truly agentic capabilities into their platforms. The rest apply the "agentic" label to rebranded assistants, chatbots and RPA workflows without changing how they function. Marketers investing in so-called agentic platforms, MarTech notes, may not be getting what they pay for.

Yet another Gartner figure points the other way: 33% of enterprise software applications will include agentic capabilities by 2028, up from less than 1% in 2024, and at least 15% of day-to-day work decisions will be made without human intervention. At roughly one routine decision for every seven made, MarTech frames that 15% as a solvable design challenge — marketers can deploy bots for the dull work while keeping agency in their own hands.

Delegate by design, not by default

Buying an agentic platform is not just another software purchase, MarTech argues. The pre-purchase phase should resemble how companies define job roles before hiring a human. That starts with choosing a management style: path-based management specifies the 20 steps of a nurture sequence, while goal-based management sets intent and guardrails — for example, lift mid-market SQL conversion 15%, hold brand sentiment at 4.0 or better, never touch an account with an open Tier 1 ticket — and lets the agent find the route.

MarTech poses three questions for every candidate task. Is it reversible? Low-risk, undoable work tolerates real autonomy; anything irreversible or trust-sensitive should stay under review. How is output reviewed before customer contact? And what happens when two agents disagree? A pricing agent and a retention agent will eventually chase competing objectives on the same account, and someone must decide in advance who wins.

Autonomy levels vary even within one product. An AI Marketing Analyst converting campaign data into insights ranks high on autonomy. A Personalization Strategist that codes, tests and deploys live campaigns deserves several rungs lower and closer scrutiny. Same folder, same afternoon — entirely different levels of autonomy.

The model requires minimal bureaucracy: one human leader can manage a small number of specialized agents grouped by outcome rather than channel. That leader's job includes validating agent logic for drift, defining context agents can inherit from one another, arbitrating conflicts, sandboxing new agents before live access and hitting the kill switch if needed. MarTech's advice: title that role and put it on the org chart.

Sequence before speed

MarTech recommends a three-step process: stabilize, standardize, then automate. Velocity on a fragmented operating model, the analysis warns, mostly means faster entropy. Marketers can hire a virtual teammate in an afternoon, but engineering the organization to absorb that teammate takes a quarter or two.

Measurement must be settled before kickoff. Agent utilization and percentage of campaigns completed are vanity metrics at best; vague business value was a top-cited reason for cancellation in Gartner's study — a measurement and strategic planning issue more often than a technology shortcoming, according to MarTech.

The upshot for 2027 planning: humans and agents are not interchangeable. An agent can report what happened and what is likely to happen next; deciding what the organization should do belongs to someone who can be held accountable. Hand that off to a directory listing, MarTech cautions, and you have outsourced strategy while believing you bought capacity. Budgeting in a binary human-or-software way may no longer work — and how many boxes on the 2027 org chart will be non-human remains the question every marketing leader now has to answer.

Source: MarTech (https://martech.org/how-to-build-a-marketing-team-with-ai-coworkers/) / Source: semrush.com (https://www.semrush.com/enterprise/seo/?utm_campaign=ic_mt_0101enterprise&utm_source=martech.org&utm_medium=referral)