AI agents need an orchestrator. That’s DAP.
“Agentic AI” is everywhere. Most of it isn’t agentic.
Vendors slap the label on chatbots and single-app wrappers. A real agent does more: it plans multi-step work, uses tools, recovers when things break, and gets the job done on your behalf. Today, the few products that do this well only handle narrow, well-structured tasks with clean APIs.
But real enterprise work isn’t clean. It runs across Salesforce, SAP, ServiceNow, and dozens of other tools. It depends on policy and approvals. It hits exceptions. An agent that works beautifully inside Salesforce can’t enforce procurement rules, can’t see your ERP, and doesn’t know when to loop in a human.
While organizations rush to deploy AI agents and assistants, productivity is slipping. Gartner found digital friction has cost workers nearly 12 points of productivity in just two years. Our own research shows employees are excited about AI but unsure how to make it part of their daily work. The result is more tools, more confusion, and often more frustration.
There is a gap between enterprise AI ambition and employee readiness. Agents can’t fix it – they have their own gaps. They often break down in messy, multi-app, policy-heavy workflows. Agent reliability falters, and autonomy is constrained where governance and risk matter the most.
The question is not whether agents are promising. It is what lets them operate at scale, safely, consistently, and with proof they are creating value.
Gartner’s 2025 Market Guide for Digital Adoption Platforms points to the answer: Digital Adoption Platforms. DAPs are already orchestrating workflows across apps, enforcing policy, and measuring outcomes. They are not another place to embed an assistant. They are the control plane where agents get orchestrated.
Expanding from guidance to orchestration
Guidance was the foundation of digital adoption. It still matters every day, and it always will. What’s changing is that guidance is no longer the whole story. The next layer is orchestration: connecting people, processes, and policy so AI can act in the right context, at the right time, and with the right safeguards.
First, people. Technology changes stick only when people change how they work. A DAP puts the right agent, guidance, and policy in front of employees while they work, making adoption intentional and measurable. Without that, agents become just another unused tool.
Second, reliability. Enterprise workflows are multi-app and policy-heavy. Agents are strong in narrow tasks. Orchestration makes them reliable at scale by providing one front door, the right context, human controls, and visibility into results. Gartner notes that many agent projects fail because value and governance are unclear. A DAP closes that gap.
Last, proof. Without evidence, AI projects fail. A DAP enforces policies, tracks outcomes, and creates the accountability leaders need. You can see this in action with SAP’s Joule action bar, powered by WalkMe. It delivers proactive AI across SAP and non-SAP systems in one experience. That is orchestration, not just intelligence.
The market signals are clear
Gartner estimates the DAP market reached $1.042 billion in 2024, growing nearly 28% year over year. This growth reflects investment in orchestration across systems, not in-app tips.
By 2027, Gartner predicts 30% of organizations will use DAP-supplied AI assistants that connect with other assistants. By 2028, 40% will use GenAI in their DAP to surface and optimize workflows automatically.
The category is in its prime. DAP is becoming the orchestration layer enterprises need to unify human and automated steps, enforce governance, and make AI’s value visible at scale.
Why agents need a DAP layer
Context. In an enterprise, work is shaped by factors agents cannot see from inside a single app: the user’s role, the approval chain in another system, the policy that governs the transaction. A DAP brings that context together. It knows who the user is, where they are in the process, and what should happen next. Without that view, agents make decisions on incomplete information. They send the wrong form to the wrong approver or suggest an action that breaks policy three systems away.
Continuity. Agents can work inside a single app, but workflows rarely stay inside those boundaries. An employee might start in Salesforce, jump to SAP for pricing, check inventory in ServiceNow, then return to Salesforce to close the deal. Right now, that means switching between different assistants that don’t talk to each other. The pressure falls on the employee to hold it all together, which is the opposite of the point. A DAP connects those steps, carries the context forward, and manages the handoffs so work moves end to end. People can focus on judgment calls, not piecing systems together.
Control. When AI touches financial approvals, customer data, or compliance workflows, intelligence is not enough. You need guardrails, audit trails, and proof. When did the agent act alone? When did it route to a human? The right DAP enforces the rules, logs the decisions, and surfaces the outcomes. That’s what lets you scale from ten users to ten thousand.
The enterprise realities AI alone won’t solve
- Adoption isn’t automatic. Employees need reinforcement in the flow of work. Without it, even the best AI tools sit unused with unrealized ROI.
- Coverage is incomplete. Many agents only work where APIs are perfect. Real workflows cross systems, policies, and exceptions.
- Sprawl is real. Different teams piloting different assistants creates a patchwork experience, weakens control, and hurts productivity.
A DAP addresses all three by unifying the entry point, orchestrating assistants across systems, and connecting everything back to adoption and outcomes.
What leaders should do now
- Make orchestration a core layer. Treat your DAP as the control plane that coordinates assistants and agents across applications. Not an add-on. Not a pilot. A foundation.
- Measure outcomes, not activity. Don’t just track clicks and logins. Tie orchestration to business results. Did the process complete? Did costs go down? Did errors drop? Use those signals to improve.
- Give employees one front door. Make it simple: one consistent way to get help or take action, no matter the system. Assistants should work together, not compete. The always-on action bar pattern, like what we built with SAP’s Joule, shows how this can look.
- Reinforce change in the flow of work. Training sessions cannot keep up with the pace of AI-driven change. Employees need support and reinforcement in the moment, while they work.
- Build trust with clear guardrails. Decide when an agent can act alone, when approvals are needed, and how exceptions get handled. Then make sure those rules are visible and followed. For financial workflows, HR data, or regulated processes, this is essential.
Looking ahead
The next phase of digital work will not be defined by a single AI assistant. It will be defined by how well organizations orchestrate the tools, processes, and people they already have.
AI will change who clicks. DAP will change how work flows. Together, they can either add to the complexity or finally deliver the simplicity we have been chasing.
Gartner now frames DAP as the orchestration layer for AI. Having worked with customers through every wave of enterprise software, I can tell you: orchestration is what separates AI projects that scale from ones that fail. Get that right, and everything else follows.