Editorial
Research Report
The State of the Digital Workplace
Workflows are back at the center of enterprise conversations. AI‑native ambitions are accelerating this shift as organizations move from pilots to embedding AI agents directly into the way work gets done.
Still early in 2026, many organizations are deploying AI agents across multi‑step workflows. However, Deloitte’s 2026 research shows that far fewer have redesigned their processes and operating models to support this shift. When agents amplify broken work and operating models, automation adds complexity and increases the cost of work.
Travel spend is a clear example. Flight delays, reroutes and geopolitical disruptions are a routine part of travel today. Yet organizations have deployed AI agents against travel workflows designed for predictability. When an employee rebooks a flight after a cancellation or missed connection, the agent flags the higher fare as noncompliant. Finance teams then supervise exceptions and overrides after the fact. Employees wait longer for reimbursement and begin to pursue workarounds as they lose confidence in the system.
Automation has added complexity precisely when judgment and compassion are most needed.
In the DEX series, we explored why fragmented workflows undermine employee experience and performance. Unifying workflow silos was the starting point. Redesigning the flow of work is the harder part. This is not a continuous improvement or automation exercise. It is an operating model decision designed through a human lens. While emerging tech and AI create new possibilities, they do not dictate how work should be redesigned.
There is no single, accepted framework yet for redesigning processes in the agentic world, and waiting for one is not a viable option. Below, I adapt established process architecture concepts to explicitly address their application in AI operating environments. Let’s jump in.
1. Reset the Starting Point
Resist the instinct to begin with tasks, tools or automation, and instead establish the work to be done and the outcome it should produce. This is process map Level 1, as defined in business process frameworks such as APQC. Each level simplifies a source of complexity that agents tend to amplify if left unexamined.
The key metrics at this level are the number of process steps and cycle time, creating a clear opportunity for work simplification.
Process map Level 2 extends traditional process mapping with insights from emerging agentic operating environments. It explicitly addresses ownership, decision rights, accountability and escalations. This must be designed, not assumed, before getting into task-level execution detail. The opportunity here is to redesign and identify human and agentic handoffs. Without this step, AI agents inherit legacy workflows and amplify existing flaws.
The key metrics at this level are the number of actors, decision points and handoffs, creating an opportunity for organizational simplification. This impact shows up directly in human and agentic org charts.
Process map Level 3 makes work executable by capturing operational context. That context makes explicit the systems involved, the data created, read or updated, and the conditions that trigger or change the flow of work. When this context is missing, workflows cannot deliver sustainable business value.
The key metrics at this level are the number of system and data dependencies, creating an opportunity for architectural simplification.
Redesign is complete at Levels 1–3. Levels 4 and beyond are situational, introduced only when execution or regulatory requirements require them. Introducing them too early is harmful and creates a false indicator of maturity.
2. Decide Before You Automate
With process maps Levels 1–3 in place, organizations can distinguish between work to eliminate, decisions to redesign and execution to automate. Deploying agents before these distinctions are clearly articulated will create workflow debt while limiting ROI. Formal BPM and workflow tools, like Camunda, Bonita and others, assume that two critical redesign steps are already complete:
-
The outcome and scope of the work are clearly defined
-
Roles, ownership and handoffs are understood
This is a sequencing issue, not a limitation of the tools. When those steps are skipped, BPM tools force teams to document noise instead of intent. The result is hard to change, brittle workflows that are rarely adopted.
A lightweight solution like Miro or Mural, when paired with standardized templates, can do this work well. Used deliberately, this approach supports:
-
Level 1 clarity on outcomes and boundaries
-
Level 2 clarity on actors, decisions and handoffs
-
Level 3 clarity on systems and execution context, without locking into automation
3. Communicate and Calibrate Through Learning
Active, simple communication is more effective than broadcasting. It drives adoption and moves redesign into use. At this stage, leaders should share learnings, adjust actions and communicate progress, without framing change as failure or rework. As this is new ground for everyone, learning becomes the signal that the organization is making progress.
This approach reflects proven practice in short, learning-led cycles of work. Communication stays close to the work, using process maps as the anchor and focusing on what is changing and why. In practice, this means:
-
Refer to the map when explaining changes
-
Highlight what moved, what simplified, what stopped
-
Avoid creating new artifacts
This lightweight model is intentional at this stage. It preserves momentum, supports learning and enables faster responses to rapidly changing conditions without introducing unnecessary overhead. Without this discipline, agentic systems scale exceptions and compound coordination costs over time.
Conference
Apr
27
Gartner Digital Workplace Summit London 2026
Register
Webinar
May
6
Retention Is a Communication Problem: New Data from 6,000+ Employees
This session gives leaders a clear diagnosis and practical path to treat retention as an infrastructure problem.
Register
Conference
Jun
2
Gartner Application Innovation & Business Solutions Summit Las Vegas 2026
Register
Webinar
On demand
Do Learning Programs Really Work? How to Turn Education Into Engagement In Healthcare
See how leaders are using learning programs to build trust with healthcare professionals and create measurable engagement.
Watch Now
As we continue this series, we will use the travel spend example to test and refine this model. Please share your own use cases so we can design, test and evolve the model together.
Editor's Note: What else should businesses consider when introducing AI agents?
- Start With the System, Not the Task: A Lean Framework for HR AI Implementation — When you automate drudgery, you optimize the past. Start with purpose and system design to create space for reinvention.
- Humans and AI Agents: Planning the Org Chart of Tomorrow — Why AI agents belong on your org chart — and how to incorporate them.
- When the AI Agent Runs Wild, Who Pays the Bill? — AI agents are making spending decisions your finance team never approved. The answer isn't better dashboards.
Learn how you can join our contributor community.