Your AI Strategy Needs More of the Right People, Not Less

Recent conversations with PE leadership can be summarized in a word: “friction.” Countless amounts of thought are dedicated to the question:

“Is now the right time to redirect budget dollars from people to the AI tech stack?”

Much of my time is consumed by this question. Within Cortado Group, we debate this on a regular basis. And (for now at least) the overall budget will remain unchanged. What is asked of employees, however, is changing in lockstep with the evolving capabilities of AI.

Your AI strategy is only as strong as your people’s willingness to use it. Earn the trust before you build the [AI tool].

— David Russell, Distinguished Innovation Fellow

Overinvest in AI

Yes. Invest budget dollars in the AI tech stack. Experiment generously. Don’t be afraid to take risks (both calculated and uncalculated). In our experience, now is the time to test the edges of this powerful innovation. Software and workflow innovation has never been this affordable, so text every major issue with an AI solution.

Even without a clear Return on Investment (ROI) in the near term, relentless curiosity will eventually yield positive results in the long term.

Invest smartly. Your proprietary data and information set should be at the core of all your innovations. This is your “Context Engine.” This is the underlying database that provides the direction of all AI applications and layers, so your team produces meaningful work. This will require careful investment, but without it you risk creating “AI Slop.”

Empathy is King

The other part of that question about headcount is more complicated. It cannot be discussed without talking about empathy and critical thinking.

I’ll start by saying that content has been commoditized. We are awash in all forms of it. We tune it out or we chop it up into meaningless sound bites. We no longer have the patience to sit with long form content.

In the past, only qualified and talented thinkers could get their ideas published. This old world made it impossible to be heard without the backing of the major publishing houses.

Online media and self-publishing disrupted that model. Turns out there are a lot of people with smart things to say. To be sure, there was more “bad content” available, but it was a small price to pay for more excellent content. It was still human-generated thought.

On the whole, we became more engaged in social discourse. Our curiosity and understanding of each other remained the most human of all our strengths: empathy.

Empathy is Endangered

AI is changing this. Authors can outsource all their writing and pass it off as their own. As the models get stronger, various AI writing “skills” and grammar tools automatically conform all writing to a universal standard. A “good writing style” according to AI, is the most efficient, non-creative expression of its view of the world. Not the view of the authors.

Invest in the (right) People

Back to the question of headcount. Invest in people as well as technology. But invest in those who are willing to use the AI tools as a compliment to their expertise.

My recommendation is to part ways with the non-thinkers and active resisters. The ones that don’t want to participate in meaningful discussion and AI usage will only slow down progress. These are the people that fear the future and losing their job instead of embracing a new future.

When measuring if your budget has been spent right now, focus less on adoption and usage metrics, and more about outcomes. Instead, improved revenue per employee can be achieved by offloading the lower-value work to AI while enabling the team to focus on forming true connections with the buyers. Monitor the quality of those connections by measuring the health of Net Revenue Retention (NRR).

This assures you have correctly allocated budget to AI tools and the team without sacrificing the quality of serving your customers.

Examples from the Field

Without a skilled human in the loop, AI tools lack effective guidance. Prompting is sloppy. AI is designed to optimize for the prompt it receives (good or bad) and not necessarily for the correct answer. This issue can become prominent with so-called vibe coding, when an un-trained employee builds an app. Bad code amplifies bad output.

For example, we discovered during a recent client engagement, a member of the sales team member used an AI tool to process a complex prospect proposal. The output looked clean, but no one reviewed it. The prospect received a quote that had to be retracted and revised. Even though the poorly written proposal was retracted, it was too late. The prospect saw the rather generous terms and accepted them to the detriment of our client. The sales rep lost their job.

The first fix is building a verification step into the workflow. Although the tool did what it was built to do; a knowledgeable human was missing from the process. No one had given AI the right direction and the result was a low margin sale and an employee firing.

Something similar occurred in the legal world. While we all enjoy a level of schadenfreude when it comes to lawyers failing, this one was of note. For those who are not familiar with the news story, a legal team filed a brief that was written by AI. The cited case law that was referenced did not exist. Needless to say, those attorneys were disbarred.

Harvey, an AI tool now used across major law firms, is a direct response to that problem. But the profession adopted earlier tools without first building quality rigor into the process. Unreviewed AI output costs people their careers and organizations their clients.

Harvard Business School and BCG studied 758 consultants using AI. Inside the range of tasks where AI performs reliably, those consultants completed 40% higher-quality work. Outside that range, on tasks where AI is less dependable, they performed 19 percentage points worse than people working without it. They didn't know they had crossed the line. Nobody told them when the tool stopped being useful and started producing liability.

Klarna is another example where the human touch . In 2023 and 2024, the company replaced 700 customer service agents with an AI assistant. The CEO initially claimed AI could handle all customer service. By May 2025, without the empathetic, human-to-human interactions, customer satisfaction plummeted, dragging retention down with it. The same CEO reversed course, saying, "Really investing in the quality of human support is the way of the future." They began rehiring. Buying a tool is a procurement decision. Redesigning how work gets done is an organizational one.

A final example comes from McKinsey's 2025 State of AI report, which found that only 21% of organizations have fundamentally redesigned workflows around AI. The other 79% use AI as an overlay on existing workflows. Most of the risk described above remains unaddressed in your portfolio companies.

How to deploy the AI-centric approach

To redesign work around AI, you need two levels of talent involved:

• the end user
• the AI manager.

End User Level

Every person using AI becomes a director of their own output. They decide what goes in, what comes out, and what needs to change. AI can do the work, but a knowledgeable person still needs to guide it and make the final judgement on the quality of the output. It requires someone who understands the domain well enough to teach AI to deliver the work at hand correctly.

What you don't want is AI producing a book report. You need AI plus a knowledgeable human to give you something insightful, directive, and executable.

AI Engineer

This is the one person who owns the tools the team uses, decides how AI gets trained and applied, and updates the approach as the market changes. Think of this as a forward-deployed engineer (ex. A member of the go-to-market team, not the IT group). In real practice, according to Lightcast’s 2025 labor market data , AI-specific job postings quadrupled from 2023 to 2024. Organizations are already hiring for this.

When designing the budget, BCG's research on AI value leaders confirms these underlying principles. The AI-focused organizations are all spending budget at specific allocations:

When deciding where to invest, the practical starting point is a workflow and issues audit. Taking a problem-first approach is always the best place to begin.

• Document what people actually do, not what job descriptions say. You cannot place human judgment where it matters if you don't know what the work actually is.
• Find the people already using AI well on the team. They become your internal proof of concept. Build the model around what they're already doing.
• Interview a small group about where work is hardest, slowest, or most error-prone.
• Map the high-stakes, client-facing touch points where common friction points cause a risk to the relationship. These are the touchpoints where the director-and-editor standard is non-negotiable.
• Apply the redesign to the place where the challenges cause the most pain, not everywhere at once.Finally, ask the redesign question: how do you rearrange these tasks so AI handles more and human attention goes where the stakes are highest?

What it looks like when the redesign works

Once the knowledgeable human in the loop becomes part of the operating model, the division of labor becomes clear quickly. AI handles volume work, while people handle judgment work. Everyone on the team knows which category a task falls into, and output gets faster and better.

Moderna is a prime example of this. They deployed ChatGPT Enterprise in 2024. Within two months, employees built 750 custom AI tools for their own workflows. The average user had 120 AI conversations per week. The legal team reached 100% adoption. This result came from intentional design based on problem solving and workflow automation. The organization gave people the context, training, and permission to rebuild how they worked. And it was successful.

Back to the original question

Is now the right time to redirect budget dollars from people to the AI tech stack?

Yes. Invest budget dollars in the AI tech stack.

Yes. Invest and retain the people that are curious and bold. Empower and promote the practitioners of empathy.

This is the talent that will remain at the very center of your strategic design and execution. A sound AI plan is only as good as these people.

As time goes on, the companies that over-invest in AI and the right people will be the ones that create new roles to run the new capabilities they have themselves created.

Sources: Microsoft/LinkedIn 2024 Work Trend Index; Dell'Acqua, McFowland, Mollick et al. (HBS/BCG, 2023); McKinsey State of AI 2025.