At Cloud Coach, we build professional services automation on Salesforce, so most of my week is spent with services leaders who are trying to deliver more with the teams they have. Almost every one of those conversations ends up on AI. The question is no longer whether to use it. Most of you already are. The question is what happens to your people while you do.

Capability Is Moving Faster Than Readiness

What AI can do keeps climbing. What has to change for any of it to matter moves at human speed: the roles people hold, the processes they follow, how decisions get made, and how much change an organization can take on at once.

Access to a powerful tool doesn't give you an organization that can use it well. Someone still has to decide what the agent should do, check what it produced, and own what happens when it's wrong. That work lands on people whose jobs were designed before any of this existed. The wider the gap between capability and readiness, the more of that weight falls on the same people.

Leaders and teams also feel that gap differently. Leaders feel the speed as market pressure, customer expectations and targets that don't wait. Teams feel it as shifting expectations, unfamiliar work, and responsibility that arrives before they feel ready for it. Eighty miles an hour in the driver's seat is not the same as eighty miles an hour in the passenger seat. The driver sees the road and knows why the car is moving fast. The passenger feels every turn without that context.

On AI, most leaders are in the driver's seat. We've seen the strategy and the timeline. Our teams feel every acceleration from the passenger seat, and what feels efficient to us can feel chaotic to them. Neither side is wrong about what they're feeling.

We've Run This Playbook Before

Most of us led teams through digital transformation, and the pattern was consistent. The technology changed first: new systems, the move to the cloud, SaaS everywhere. The work shifted around it, with new processes, new roles and more tools to keep current. Then people absorbed whatever the change demanded, usually on top of the job they already had. Training happened in the margins, workarounds piled up, and the load on people kept growing long after the project was called a success.

In McKinsey's 2018 global survey on digital transformation, only 16% of respondents said their transformations had both improved performance and equipped the organization to sustain those improvements over the long term.

My concern is that we're about to run the same playbook with AI, only faster, and this time the work is changing at the level of judgment and responsibility rather than screens and clicks. So the question isn't whether to move. It's how we decide whether the change is working.

A Shared Language for What Gets Better

Most organizations measure AI by features deployed or hours saved. That's the organization's scoreboard. The person doing the work asks different questions: 

• Is my work better? 

• Can I still do it well? 

• Can I sustain it? 

The customer asks whether the service they get is better. When those views never meet, every AI conversation turns into a negotiation between people reading different scoreboards.

The most useful language I've found for where they meet comes from Graeme Smith's essay "The Human Operating Envelope." In plain terms, and as we apply it at Cloud Coach: net human gain means people and customers are better off after you account for the new demands AI creates.

That last clause matters most. Smith gives the example of an agent doing three hours of work while someone is at lunch. The output is waiting when they get back. So are the checks and decisions it requires. AI produces the work faster, but people still have to direct it, verify it, correct it and fit it into everything else. Those demands are real, and they decide whether the apparent gain turns into something better for people.

So recovered time deserves an honest count. Take the human hours a result needed before AI, and compare them with all the human hours it needs now for the same accepted result: briefing the agent, checking and correcting its work, the effort passed to colleagues, and the setup and support over time. The difference is the time you actually got back.

Time is only one part of it. The larger question is what improves in people's working lives:

• Do they have greater capability and control over the work?

• Is the workload sustainable?

• Is the team engaged and bought in, or is the change being done to them?

• Does all of that show up as better outcomes for customers?

A faster deliverable can coexist with weaker understanding or greater dependence, so we look for evidence on each of these instead of rolling them into one score. What we're trying to create is capacity for growth: room to serve more customers well, develop our people and take on work we couldn't support before, in a way the team can sustain.

Start With What the Role Exists to Deliver

The easy move is to leave every job description as it is, give everyone an AI license and call that a strategy. A better starting point is to ask what a role exists to deliver, and which parts of it a person should still do now that agents can handle much of the production work.

Take a project manager in a services organization. The role exists so that projects finish on time and on budget, with a client who wants to work with you again. Break the role into individual tasks and make a choice for each one:

• Automate: the agent performs the task and a person checks it. Status report drafts, meeting notes and action items, timesheet reminders, risk and issue log updates.

• Augment: the agent drafts and the person edits, judges and owns the result. Project plans and schedules, risk assessments, change requests, resource forecasts.

• Human-only: client relationships, escalations and difficult conversations, scope and commercial decisions, and accountability for the outcome.

Done this way, you end up with two job descriptions instead of one: one for the person and one for the agents. The agent description spells out the work, its boundaries and where it hands off to a person.

The human job changes too. It now includes directing and evaluating the agents’ work, which means judging and improving work the person no longer does themselves. That responsibility will reach every delivery team, and most of the people taking it on have never managed anyone.

Everyone Becomes an Accidental Manager

Rowland Manthorpe has written about the disappearance of the secretary, and it explains what comes next. A manager once dictated a memo, and someone else typed it, copied it, distributed it and filed it. Personal computers put that work in the hands of the person who needed it done. Over time we all took on our own administration: our own documents, calendars, travel and expenses.

Delegating to agents runs the other direction. We still have to explain the outcome, provide context, judge what comes back, handle exceptions and decide what happens next. Those are management responsibilities. As people start managing work that agents perform, management becomes part of their job, including for people who have never managed anyone.

We can see that responsibility coming, so we can prepare people for it. Managing delegated work is a skill, and most of the people inheriting it have never been trained to do it. They need practice defining a useful outcome: "Fill out this spreadsheet." and "I'm trying to reduce what we spend." are very different instructions. They need practice giving enough context, because handing over a task with less context than you had and expecting a better result is how delegation fails. They need to recognize a good or bad result when it comes back, and know when to intervene. That's management, not AI engineering, and people get better at it with repetition and feedback.

Leadership's job is to make that possible. People need:

• A defined outcome, so they know what success looks like.

• Authority to accept, reject or stop agent work.

• Preparation: the skills and context to judge the result.

• Capacity: protected time for oversight within the role, not extra hours on top of it.

This is also why outcomes matter more than static processes. The tools will keep changing. Someone who understands what success requires can adapt as the method changes. Someone who only learned the steps has to be retrained every time.

People Don't Experience Strategy. They Experience Tempo.

At the start of this year our company theme was efficient and effective: remove drag, increase precision, move as one. By midyear it had become "no time for caution," because the market accelerated and caution had become the risk. I still believe that, and we still have to move quickly. What I've learned since is that speed alone isn't the answer. If the work moves fast and the responsibilities and preparation don't move with it, you get the same gap between capability and readiness.

Tempo is the rhythm of change across a company. People experience it through shifting priorities, new initiatives, the speed of decisions and how much they have to absorb at once.

Too slow and you miss the window. Too erratic and you create chaos. When those things are disconnected, even a sound strategy feels chaotic. Leaders set the tempo through how they sequence changes, communicate priorities, make decisions, and give teams room to finish, learn and adapt.

How We See It

AI is moving faster than organizations can absorb it, and leaders decide whether that speed turns into something better for people. Three commitments follow from that:

1. Judge AI's return by its benefit to people and customers. Look for better outcomes, greater capability and real capacity for growth, and count the human work it takes to get there.

2. Prepare people to manage agents. Coach delegation, judgment and intervention, and give people the authority, preparation, and time the job requires.

3. Set a tempo people can sustain. Move quickly, and help teams absorb change and keep delivering while they do.

The benefit to people doesn't happen on its own. It depends on the conditions leaders create around them. Get those right, and AI creates capacity for growth that your team can sustain.

Frank Nardi presented "Net Human Gain: Preparing Your Services Team for AI Transformation" at TSIA Technology & Services World (TSW) 2026 in Las Vegas.

Sources Cited

McKinsey & Company, “Unlocking Success in Digital Transformations,” McKinsey Global Survey, October 2018 (online survey of 1,793 participants, January 2018). https://www.mckinsey.com/capabilities/people-and-organization/our-insights/unlocking-success-in-digital-transformations

Graeme Smith, “The Human Operating Envelope,” SIGNAL 006 (Signal Intelligence Briefing), July 29, 2026. https://thisisgraeme.me/2026/07/29/human-operating-envelope-signal-006/

Graeme Smith, “What Is The Human Operating Envelope?” (companion field note), July 29, 2026. https://thisisgraeme.me/2026/07/29/the-human-operating-envelope/

Rowland Manthorpe, “The Case of the Disappearing Secretary,” Rowland’s Newsletter, March 1, 2026. https://rowlandmanthorpe.substack.com/p/the-case-of-the-disappearing-secretary