
2,000 Agents and No Orchestrator
This is Part 4 of a five-part series. Part 1 exposed the failure numbers. Part 2 diagnosed root causes. Part 3 revealed the winning architecture. This week: the next wave is coming, and it will separate the prepared from the destroyed.
Enterprise AI Research That Actually Helps You Win. Get weekly research, benchmarks, and practical insights on how leading companies turn AI into measurable ROI.
Every enterprise vendor presentation in 2026 leads with agents. Every roadmap features them. Every pitch deck has the word "agentic" on slide three.
The opportunity is real. Gartner predicts that 40% of enterprise applications will include task-specific AI agents by end of 2026 [7], up from less than 5% today. Their best-case scenario projects agentic AI could drive 30% of enterprise software revenue by 2035 — more than $450 billion. This is not incremental. This is a platform shift.
But here's what the vendor presentations leave out.
Gartner simultaneously predicts that over 40% of agentic AI projects will be cancelled by end of 2027 [8]. Escalating costs. Unclear business value. Inadequate governance. Sound familiar? It should. It's the exact same pattern we saw with generative AI — except agents have the added complexity of autonomy. They don't just answer questions. They take actions. And when ungoverned agents take wrong actions at scale, the consequences are immediate and expensive.
Perhaps most damningly, Gartner estimates that only about 130 of the thousands of claimed agentic AI vendors are legitimate [8]. The rest are "agent washing" — rebranding existing chatbots, RPA bots, and workflow tools as AI agents without genuine agentic capabilities. The enterprise AI market's credibility problem is about to get dramatically worse.
Here's the insight most organisations are missing: agents need a platform to run on.
One enterprise CTO [21] reported scaling from a handful of pilot agents to nearly 2,000 AI agent instances across 40+ types. That proliferation happened fast. Without intentional architecture, it creates exactly the kind of unmanageable sprawl that kills enterprise technology programmes: agents making decisions nobody can trace, taking actions nobody authorised, consuming compute resources nobody budgeted, and when they go wrong — no kill switch, no rollback, no audit trail.
The pattern from every previous enterprise technology wave applies here with even greater force. Cloud created server sprawl. Mobile created app sprawl. SaaS created vendor sprawl. Agentic AI will create agent sprawl. And the organisations without an orchestration platform will learn every painful lesson of the past two decades again — faster, more expensively, and with higher regulatory risk.
Forrester predicts that 30% of enterprise app vendors will launch MCP (Model Context Protocol) servers in 2026 [10], creating an open interoperability standard for how agents communicate across systems. This is the moment when platform architecture becomes decisive. The organisations running agents on a governed orchestration platform with standards-ready APIs will plug into this ecosystem seamlessly. Everyone else faces an integration tax that compounds with every new agent.
Now let's talk about what AI agents mean for people.
The headlines scream job destruction. The data says something different.
McKinsey's 2025 survey [2] shows mixed workforce expectations: 43% of enterprises predict no change in headcount, 32% expect decreases, 13% expect increases. But EY's US AI Pulse Survey [16] found that only 17% of employers actually benefiting from AI are reducing staff. Instead, 38% are reinvesting productivity gains into upskilling existing employees.
PwC's extraordinary 2025 Global AI Jobs Barometer [15] — analysing nearly one billion job ads across six continents — concludes that AI makes people more valuable, not less. Workers with AI skills command a growing wage premium. Companies using AI purely for headcount reduction may capture short-term savings while missing the much larger prize of new market creation.
But there's a massive adoption gap. WalkMe [13] found that enterprises deploy an average of 200 AI tools — but only 28% of employees know how to use them. The tools exist. The training doesn't. And the platform that connects tools to actual daily workflows — making AI invisible inside the process rather than another app to learn — is missing.
This is where the human cost of the platform gap becomes real. Knowledge workers are drowning in tools they didn't ask for, can't use effectively, and don't trust. The solution isn't more training on more tools. It's embedding AI into workflows people already use, on a platform that handles orchestration invisibly.
Forrester predicts the hype correction is imminent: enterprises will defer 25% of planned AI spending into 2027 [11] as ROI scrutiny intensifies. The correction will reward disciplined organisations with governed platforms — and punish those still running unfocused experiments.
The window between hype and reckoning is closing. Next week: what the 6% actually do differently — and how to join them.
Sources:
[2] McKinsey, Nov 2025 → https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai
[7] Gartner, Aug 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
[8] Gartner, Jun 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
[10] Forrester, Nov 2025 → https://www.forrester.com/blogs/predictions-2026-ai-agents-changing-business-models-and-workplace-culture-impact-enterprise-software/
[11] Forrester via Jones Walker → https://www.joneswalker.com/en/insights/blogs/ai-law-blog/ten-ai-predictions-for-2026-what-leading-analysts-say-legal-teams-should-expect.html
[13] WalkMe, "SODA 2025" → https://www.walkme.com/state-of-digital-adoption/
[15] PwC, "2025 Global AI Jobs Barometer" → https://www.pwc.com/gx/en/services/ai/ai-jobs-barometer.html
[16] EY via HRD America → https://www.hcamag.com/us/specialization/change-management/just-17-of-employers-reducing-headcount-amid-ai-gains/560066
[21] CIO.com, Feb 2026 → https://www.cio.com/article/4125864/agentic-ai-isnt-about-the-agents-its-about-the-platform.html
Ready to Turn AI Into Real Business Value? PullStream helps organisations deploy AI through workflow automation, integration, and governance that actually scales.