AI can turn unclear priorities, weak ownership and poor decisions into expensive mistakes faster than leaders can contain them.
AGI, or artificial general intelligence, describes AI that could handle almost any thinking task a capable person can. Google DeepMind's March 2026 paper proposes how to measure progress toward it. You can't wait for that debate to settle before closing your management gaps.
Whether you call it AGI or highly capable AI, the management fundamentals haven't changed: clear outcomes, named owners and the authority to deliver. What's changed is the speed. Management weaknesses you've tolerated for years now compound in days.
Here's a pattern I keep seeing. Someone two levels below the CEO builds a working tool in days. It pulls from three systems, flags accounts at risk of leaving and drafts the outreach. A year ago, that was a quarter of IT's roadmap. Everyone's impressed.
Then I ask: What reduction in customer churn are we targeting, and by when? And who's accountable for delivering that result and keeping the tool reliable as the business changes?
The room goes quiet.
The risk is that our ability to build has outpaced our discipline in defining success and assigning ownership.
In a study of 5,172 customer support agents at one company, AI assistance increased issues resolved per hour by 15% on average, with larger gains for less experienced workers.
Consider a sales team using AI to prepare account briefings before customer meetings. It brings together prior conversations, open issues and customer priorities. The goal is less preparation time and more productive customer conversations.
Give your people tools, training and clear expectations to find opportunities in their daily work. Then measure whether they're saving time, improving quality and delivering better results.
Access to knowledge doesn't settle what to build, why it matters or what happens after launch.
I've watched this pattern for three decades, from mainframes to SaaS and AI. We launch a solution people need, but it never becomes part of how they work. AI shortens development time. Leaders still have to provide the training, set expectations and change the processes that turn a useful tool into a business result.
AI makes that failure cheaper to commit and easier to repeat. Companies fill with orphan solutions: tools without clear outcomes, owners or plans for changing models, rising costs and security problems.
Each one looked like progress. Together, they're a liability with a login page.
Give people room to experiment. Before the business relies on what they've built, require three sentences:
We're solving this problem for this person or team.
Success means this metric reaches this target by this date.
This person owns the outcome and is responsible for keeping the tool useful, reliable and within budget.
Managing in the AGI era means delegating work without giving away accountability. Define the outcome. Have the owner test the full workflow and demonstrate consistent results before expanding what the tool can do independently. Set clear boundaries for decisions, human review and escalation.
Your job is to give people the authority to use AI, the standards to judge its work, and the responsibility to act when it falls short. A successful demonstration starts that responsibility. It doesn't finish it.
Every person should know where the company is going, why their work matters, what success looks like and which decisions they can make.
Ask each person to explain success in their role this quarter in one sentence. Then ask how it supports the company's priorities.
Ten people building quickly toward conflicting priorities create more work to reconcile. Clear outcomes give people a reason and direction to move faster.
Across the companies I've helped scale, private equity-backed and venture-backed, one thing never changed through downturns, integrations and technology shifts. People follow leaders they trust.
I come back to HEAR: honesty, empathy, accountability and respect.
Honesty: Tell people what AI changes about their jobs, including work that may disappear. Be clear about what you know and what you haven't decided.
Empathy: Saving someone hours can make them wonder whether they're still needed. Acknowledge that concern. Give them training and a credible path to more valuable work.
Accountability: AI can sound confident and still be wrong. If a number leaves with your name on it, you own it. "The model said so" doesn't transfer responsibility.
Respect: Respect people's time. Don't give someone a tool that saves two hours, then fill those hours with reports nobody uses.
Give people decision rights, a budget they can manage and clear escalation limits. Match their authority to the outcome they're expected to deliver. Requiring your approval for every decision makes you the bottleneck.
Cost savings matter, but so does what your people can accomplish with the capacity AI creates. Before asking how many roles you can remove, ask what your team can deliver that was out of reach.
Heroes don't scale. Systems do. That system depends on people who know what success looks like, own it by name and have the authority to deliver it.
First, pick an AI tool your team already relies on. Write three sentences: the problem and audience, the measurable outcome, and the person responsible for the result and ongoing operation. Close any gaps.
Second, ask five people what success means in their role this quarter and how it supports company priorities. Different roles should have different answers. If people can't explain the connection, that's your management work.
Third, choose one leader who regularly waits for your approval. Agree on a measurable outcome, record current performance and give them authority over specific decisions, such as approving customer credits up to a set amount. Define when they must involve you. Review progress weekly. After 90 days, assess whether decisions are faster and business results have improved.
The machines keep getting more capable. Give people the clarity, trust and authority to turn that capability into results.
Find out who your company waits for. Take the free one-minute Hero Test →
Related KeyDelta Services

Russ Reeder
Founder & CEO, KeyDelta | Forbes Technology Council
Author of Heroes Don't Scale and an expert in operational excellence and AI transformation for PE-backed companies. He has spent most of his career scaling companies, including private equity-backed and venture-backed businesses, often as president or CEO, among them Media Temple (a GoDaddy company), OVHcloud U.S., Infrascale, Netrix Global, and XTIUM. Founder of Rightsline (Disney, Netflix, Amazon MGM Studios, Merck). Forbes Technology Council member. HBS Executive Education.
Want to discuss these ideas?
If your team is navigating execution challenges, we should talk.
Not ready to talk? Score your AI readiness in two minutes →