Redesign Work Before You Add More AI Agents

1. AI Everywhere. Workflows Untouched.

on Agentic AI workflow enablement across enterprise teams, I kept seeing the same problem. Teams told me they were excited about AI. Then I watched their day-to-day: a workflow pulling from different data sources, multiple rounds of Excel copy-paste, and manual handoffs before anything reached a checkpoint. The most important intelligence rarely made it into any system. It stayed in people’s heads, undocumented, scattered across emails, slide decks, and chat threads. 

People were using AI, but the work itself still ran the old way.

I kept hearing the same proposed solution too. When teams talk about AI, someone says: let’s put a nice chatbot on top of our files and data, and everything will work out. I have spent weeks sitting down with teams to do a deep dive on their current processes. It is slow, sometimes quite painful, and it is the necessary step that unlocks the AI value. And I am certain these cases are far from unique.

People love to discuss AI adoption at a high level. But what I want to share may sound hard to hear for a lot of people. AI becomes business value only when it reaches your products and business processes. The truth is that before you introduce more AI tools and agents, you need to redesign and improve the workflow.

Recent research points to the same conclusion. McKinsey’s “Talent to Value” work shows that AI value is increasingly created by coordinated systems of humans and agents, and it cites Johnson & Johnson’s nearly 900 GenAI use cases, where 80% of value came from only 10% to 15% of initiatives. BCG’s 2026 AI Radar adds the CEO lens: companies expect AI spending to roughly double in 2026, and nearly all CEOs believe AI agents will produce measurable returns this year. Microsoft’s 2026 Work Trend Index brings it down to how work gets done: its most advanced AI users use agents for multi-step workflows, rethink workflows, and create shared AI standards for their teams.

Your AI strategy should no longer start with a list of siloed use cases. Instead of scaling AI tool access, it should start with a value map, then identify the key workflows, the human judgment required, the agents that can improve execution, and the performance system that proves whether the whole setup is working.

2. Start with Your Business Value

The right starting point of an AI enablement strategy is identifying where AI can create disproportionate advantage in cost, growth, innovation, or business model expansion.

BCG’s 2026 AI Radar shows companies expect AI spending to roughly double in 2026, and nearly three quarters of CEOs say they are the main AI decision maker in their organization. The same research says nearly all CEOs believe AI agents will produce measurable returns in 2026.

Johnson & Johnson’s experience is the strongest case here. Broad experimentation across nearly 900 use cases helped the company learn, but the impact came from narrowing resources toward the 10% to 15% of initiatives producing most of the value.

The right starting point of an AI enablement strategy is identifying where AI can create disproportionate advantage in cost, growth, innovation, or business model expansion.

1. Where AI can reduce the cost.
2. Where AI can improve revenue, profit, or customer experience.
3. Where AI can support a new product, service, or business model.

Then prioritize by asking the team one question: which 10% of our AI work could create 80% of the business value?

3. Redesign the Work: From Individual Role to Human-Agent System

The old question was: who is the right person for this role? The better question is: which parts of this workflow should be owned by people, which parts should be handled by agents, and where does human judgment need to stay in control?

McKinsey’s agents for growth article shows organizations create more value when agents improve end-to-end processes instead of isolated tasks. Microsoft’s 2026 WTI reaches a similar conclusion: advanced AI users are already using agents for multi-step workflows and rethinking how work gets done.

A customer service agent that only helps employees write faster replies improves one step. A better workflow predicts issues, triggers outreach, routes exceptions to humans, and closes the loop with personalized resolution. That is a system design decision, and no AI tool deployment makes it for you.

4. Redefine Talent: AI Super Users Are Workflow Designers

The most valuable employees in the AI workplace are not always the people writing the most prompts. They are the people who can focus on the right problem, explain the current process, identify weak handoffs, implement and test AI solutions, and make the improved work scalable for others.

PwC’s 2026 Global AI Jobs Barometer shows why this talent standard is rising. Jobs requiring AI skills are growing almost eight times faster than the overall job market, and the average wage premium for AI skills has reached 62%.

So find your existing AI super users. Give them a workflow mandate instead of a side project. Ask them to document how they work, where AI helped, where it failed, and what other teams can reuse.

The opportunity is significantly larger than individual productivity. The best AI super users can help the company redesign how work gets done.

5. Educate the Executive Team Before Scaling More AI Agents

While senior leaders sponsor a lot of AI pilots, they might not have an aligned approach to decide which work deserves AI agents, which talent should be reassigned, and which business metrics they need to measure impact.

This is now a CEO-level operating issue. BCG’s AI Radar shows 72% of CEOs say they are the main AI decision maker, and half believe their job depends on getting AI right.

 You need to keep AI from becoming a scattered innovation portfolio without control.

1. Which AI projects are producing measurable business value?
2. Which projects should be stopped?
3. Which workflows need redesign before adding more tools?
4. Which leaders own the business outcome?
5. Which risks require governance, audit, or human review?

6. Measure The Business Impact

AI agent performance should be evaluated on decision quality, reliability, speed, and cost. Humans should be evaluated on business impact, AI workflow improvement, ethical use, and cross-team collaboration.

The governance gap is real. Deloitte’s 2026 State of AI in the Enterprise research found that only 21% of surveyed organizations have a mature governance model for autonomous AI agents, while about 80% lack mature capabilities such as decision boundaries, real-time monitoring, and audit trails.

An AI agent may make one task faster while slowing down review, approval, or customer resolution. The success metric is the full workflow outcome: faster decisions, fewer errors, lower cost, better customer experience, and stronger human accountability.

Use three layers of measurement:

1. AI agent metrics: accuracy, reliability, speed, cost, escalation quality.
2. Human metrics: business judgment, workflow improvement, ethical use, collaboration.
3. Business metrics: cycle time, decision quality, customer impact, cost-to-serve, continuous improvement.

7. Final Thoughts

Before your next AI review, ask one question: are we buying more AI, or are we redesigning the work that can produce the better business result?

First, stop asking for more AI use cases and start identifying the few value pools that deserve investment. Second, redesign one critical workflow by deciding what humans should own, what agents should handle, and where human review is required. Third, update the management system so AI value is measured through business results, workflow quality, and responsible execution.

This is how you protect time and budget while keeping AI adoption alive. It is also the test of whether your AI strategy is ready for business execution.