This McKinsey report, Agents, Robots, and Us (Jan 2026) was deserving of two blog posts.
The first one here segregates the different combinations of work that will be done by humans, agents, and robots (and combinations therein). TLDR; change is coming. Be thoughtful how you stack your skills, by emphasizing flexibility and continuous learning.
The second one is below:
With GenAI, we’re thinking too small
That’s already a crazy claim because AI capex is trillions of dollars, hiring for many jobs seem to be frozen, and it’s upsetting all industries from accounting, consulting, education, travel, and video editing. It’s ubiquitous.
Why is the scope bigger than we think?
So far, a lot of GenAI B2C use-cases (e.g., using ChatGPT to craft emails, brainstorm dinner ideas, translate things) were task oriented. I type in XYZ, and the machine does ABC. Google – in plain language, customized, personable – on steroids. Even GenAI B2B use-cases (e.g., reducing human demand for monotonous, algorithmic, “do it again” work) were focused at the task level. Individual tasks. Team tasks. Department tasks.
Let’s think bigger.
“Digital employees”
Yes, that’s what forward-thinking HR departments call agents. Digital agents are able to take action based on guidelines, algorithms that users have set. This is way more than answering questions and drafting emails.
Do you remember how amazing RPA (robotic process automation) revolution seemed 5 years ago? In the old days (circa 2018), you wire up a bunch of different apps and give them very specific IFTTT (if this, then that) directions. Zapier built an entire business around it.
Agents = different level. Now, agents can evaluate options and select among the choices. So, instead of hard programming a IFTTT workflow, you can have digital agents that “think and decide” and “improve” over time.
Reimagined workflows
Some companies report NOT getting significant gains from GenAI initiatives simply because the application is voluntary, ad hoc, unstructured, or simply juvenile. Like dropping a food processor in the corner of a commercial kitchen to see what happens. No, the real, structural, strategic win is when the work is reimagined (not with GenAI as a voluntary add-in to your browser), but fundamentally how we do the work.
McKinsey took a look at 80 different use cases. As a simple example, a technology company had thousands of smaller accounts which were not getting attention from their inside sales force (of course, the accounts were small). With GenAI, the company had 4 different agents: 1) prioritized the accounts by public & proprietary account data 2) outreach to the accounts to qualify them 3) managed follow up and stratified the leads (good, okay, bad) 4) schedule contact with a human employee. Results? Saved 30-50% of time, and raised revenue from these small-accounts 7-12%. Those are big numbers, right?
Managers and specialists are increasingly acting as orchestrators and validators rather than executors, while domain experts such as data analysts, underwriters, and engineers partner with agents that perform initial analysis or generate draft outputs. – McKinsey
Beyond this, we could also have agents that 5) predictively see which accounts might leave, switch 6) coach inside salespeople on ways to do a better job next time. The report has even more examples for pharmaceutical research, IT modernization, and other use-cases.
Not incremental change
This takes guts. This is not a 10% change in the way this work is done. It’s a 100% change. If you have Harry Potter magic to do mundane, repetitive, (or difficult and novel) work, why wouldn’t you use it?
This becomes a crazy mix of a) humans will do b) agents and robots will do c) humans & agents & robots will coordinate together. This is an example of YES, AND.
Courage and leadership
From my 30+ years of experience in corporate, I’ve come to a simple & somewhat obvious observation: people don’t like change. Change is tough. Companies gain massive inertia. There are norms, “ways of working”, and entitlement. Change is glacial.
So McKinsey offers these questions to ask:
- Are you reimagining your business for future value?
- Are you leading AI as a core business transformation?
- Are you building a culture of experimentation and learning?
- Are you building trust and ensuring safety?
- Are you equipping your managers to lead teams of people, agents, and robots?
- Are you preparing your workers for new skills and roles?
Damn, I believe that most of the organizations that I know would score a 2 out of 6 on that list. Grrr.
So what?
If you buy the argument above, here’s a thought on how you could use GenAI with your work.
Download the McKinsey report mentioned above, and upload into ChatGPT to give it some context for what you are doing, and prime the conversation.
1) Diagnose your current workflow
This is the most difficult and painful step. You need to provide a fair amount of detail of the 50 steps you take. . . in fact, you probably don’t even recognize how much minutiae you are doing. When I was a consultant, we used to call this “value stream mapping” and it was time-consuming, invasive, but also enlightening.
Even if you don’t automate ANYTHING, it’s still super useful to catalog, dissect, and consider what of the activities you are doing that are valuable vs. routine). In other words, give the boring stuff to the agent.
Copy and paste this into ChatGPT:
“I want to improve my _______ workflow. Please help me:
1) List the end-to-end steps involved, from start to finish.
2) Identify which steps are: a) repetitive or rules-based, b) judgment-heavy or creative, c) coordination or monitoring focused.
3) Highlight the biggest time sinks and friction points. Assume no automation yet. Be concrete.”
2) Redesign the activities
So here is where the rubber meets the road. We will be asking ChatGPT what are areas that agents can do this work just as good (or better than I can) at pennies on the dollar? Basically, what work should I give away?
Copy and paste this into ChatGPT:
“Take my _______ workflow and redesign it assuming:
– Humans focus on judgment, framing, decisions, and exceptions.
– AI agents handle analysis, coordination, monitoring, and first drafts.
For each step, specify: A) Human role B) AI agent role C) Checkpoints where humans stay in control”
3) Stay open-minded
There is a GOOD chance that you won’t automate your workflow. I didn’t. However, it’s just as instructive to figure out what we should do more of and WHAT WE SHOULD DO LESS OF:
Copy and paste this into ChatGPT:
“Based on my _______ workflow, analyze:
1) Which of my current skills are most leveraged in a human-AI partnership.
2) Which skills are becoming less valuable or more automatable.
3) Which new skills I should develop (e.g., AI oversight, prompt design, sense-making).
Propose a simple 60–90 day skill-upgrade plan.”
4) Create a system
Okay, this is for you nerds out there. Extra credit if you create a system for this stuff:
Copy and paste this into ChatGPT:
“Help me add feedback loops to my _______ workflow. Specifically:
– What signals should I track (speed, quality, outcomes, errors, user response)?
– How can AI agents summarize patterns and surface insights?
– How should I adjust decisions or workflow steps based on those signals?
Design this as a lightweight, repeatable system.”
Fix the process first
When I was at Deloitte (read 2005-2009), we used to say that you should always fix your process first. If you take a bad process and speed it up with technology, all you get is a faster death.
This is about orchestration, not just acceleration. We’re all going to have to learn how to dance with the agents.