Generalists Own the Interface. Specialists Own the Outcome.

The generalists have arrived

In 2026, xAI launched Grok Bot, a general-purpose agent designed to act as an "AI teammate". According to xAI's product page, it gets its own cloud computer, can sign into a user's existing tools, use them as a person would and return with finished work [1]. It's sold across roles such as project management, design, go-to-market and support, and is bundled into consumer and team subscriptions, with additional usage charged per token [1].

Grok Bot isn't alone. The largest AI labs are all moving from chat assistants towards agents that act. And the underlying trend is broad: Gartner expects 40% of enterprise applications to include task-specific AI agents by the end of 2026, up from under 5% in 2025 [2].

For businesses, this is good news. General agents can take on a huge range of routine work, from updating spreadsheets to drafting reports and operating software dashboards.

It also raises a hard question for anyone building specialist software: if a general agent can operate any tool, what's left for specialists?

The warning sign: projects that fail

The same research that shows agents spreading also shows many of them failing. Gartner predicts that more than 40% of agentic AI projects will be cancelled by the end of 2027, citing escalating costs, unclear business value and inadequate risk controls [3].

Gartner also warned of widespread "agent washing", existing products rebranded as agents without real autonomous capability, and estimated that only around 130 of the thousands of vendors claiming agentic capabilities were real [3].

The common thread in these failures isn't that the agents can't do tasks. It's that doing tasks isn't the same as delivering value. An agent that operates a tool faithfully but produces no measurable business result is a cost, not an asset.

Why marketing exposes the difference

Marketing is a clear test case, because the gap between doing tasks and delivering outcomes is so large.

A general agent can log into an advertising dashboard, set up a campaign, adjust a budget or write ad copy. For a small business running simple campaigns on a single platform, that may be good enough.

But much of the value, and much of the waste, in advertising sits in places a general agent operating the same dashboards as a human can't easily reach:

• The supply chain. Independent research found that only about 51% of programmatic spend reached publishers, with 15% of spend untraceable [4]. Fixing that means changing how and where media is bought, not just operating the buying tools.
• Fraud and attribution. AppsFlyer found a 36x fraud gap between affiliate traffic and self-reporting networks, and that 52% of fraudulent installs were organic users the brand would have got anyway [5]. Spotting that takes pattern data across many campaigns and brands.
• Outcome measurement. The ANA found that top-performing advertisers turn 54% of programmatic spend into qualified impressions, while others manage 32% [6]. Closing that gap depends on what you optimise for, not on how quickly you click.

These are problems of data, supply and incentives, not of operating software.

Where specialists win

In a world of capable general agents, specialists need an advantage a general model can't simply acquire. In marketing, four stand out.

1. Supply

Relationships with publishers, direct-app inventory and device makers are negotiated and contracted by people. They take time to build and aren't something a general agent brings with it.

2. Outcome data

Knowing which publishers deliver valuable users, how quickly creatives decay, what fair prices are for direct inventory and what fraud looks like requires data built up from real spend across many brands. That data only builds up where money actually moves.

3. Accountability

A general agent is usually paid for access and usage, whether or not a campaign makes money [1]. A specialist can take on part of the risk by pricing on results, for example a share of the return on ad spend (ROAS) it delivers. That's a different promise: not "we'll do the work" but "we'll deliver the outcome."

4. Neutrality

The large ad platforms already offer their own automation that optimises for outcomes. But each platform's automation works inside its own inventory and on its own terms. A specialist working across platforms can move budget to wherever it performs best, including supply the platforms don't sell.

The bigger competitor isn't a lab

Seen this way, the most direct competition for specialist marketing agents isn't a general-purpose AI lab. It's the automation built into the largest ad platforms themselves.

Those tools are powerful and increasingly outcome-focused. But they optimise within one platform's boundaries. A brand that relies entirely on them has no neutral view across channels, and no access to the direct supply outside them. For a market where much of the waste sits in the gaps between platforms [4][5], neutrality is a real advantage.

Generalists and specialists together

The most likely future isn't generalists or specialists. It's both, working in layers.

• General agents become the interface. People describe what they want in plain language to the assistant they already use.
• Specialist agents do the specialist work underneath. When the task is "grow my app's paying users at a 2x ROAS", the general agent hands it to a system built for exactly that.

There's a familiar parallel in software. Most apps don't build their own payment systems; they call a specialist provider that handles payments, fraud and compliance. In the same way, general agents are unlikely to rebuild performance marketing from scratch. They'll call the specialists that already have the supply, the data and the accountability.

In that model, the growth of general agents helps specialists. Every new general agent is another route through which specialist capabilities can be delivered.

What this means for buyers

When evaluating any AI agent for marketing, general or specialist, four questions cut through the noise:

1. What outcome is it accountable for? Tasks completed, or revenue, retention and ROAS?
2. How is it paid? Per seat and per token, or linked to results?
3. What can it reach? Only the dashboards you already use, or better supply and data?
4. What guardrails does it have? Budget limits, brand rules and approval points you control?

The takeaway

General-purpose agents are a genuine breakthrough, and they'll handle a vast amount of everyday work. But the evidence on failed agent projects [3] is a reminder that automation without accountability doesn't create value. In marketing, where the biggest losses sit in supply chains, fraud and misaligned incentives, value will go to specialists that own the outcome, and increasingly, to specialists that general agents call.

Generalists own the interface. Specialists own the outcome.

---

References

[1] xAI. Grok Bot (product page). 2026. https://x.ai/bot

[2] Gartner. 40% of enterprise apps will feature task-specific AI agents by 2026. August 2025. https://www.gartner.com/en/newsroom/press-releases/2025-08-26-gartner-predicts-40-percent-of-enterprise-apps-will-feature-task-specific-ai-agents-by-2026-up-from-less-than-5-percent-in-2025

[3] Gartner. Over 40% of agentic AI projects will be canceled by end of 2027. June 2025. https://www.gartner.com/en/newsroom/press-releases/2025-06-25-gartner-predicts-over-40-percent-of-agentic-ai-projects-will-be-canceled-by-end-of-2027

[4] ISBA / PwC. Programmatic Supply Chain Transparency Study, executive summary. December 2020. https://www.isba.org.uk/system/files/media/documents/2020-12/executive-summary-programmatic-supply-chain-transparency-study.pdf

[5] AppsFlyer. State of Fraud for Marketers, 2026 edition. April 2026. https://www.appsflyer.com/resources/reports/state-fraud-marketers/

[6] ANA / TAG TrustNet. Programmatic Transparency Benchmark: Q1 2026. May 2026. https://www.ana.net/content/show/id/pr-2026-05-programmatic