How to Make AI Agents Your Personal Assistant for Better Workforce Decisions

In our latest webinar, Haris Ikram, Co-Founder & CEO of CandorIQ, sat down with Jeet Mukherji, Co-Founder of Kinfolk, and Dave Carhart, People Partner at People Catalyst, to skip the AI hype and talk about what they’re actually doing with it: the agents they’ve built, the workflows that are working, where humans still need to step in, and how teams can get better at AI without feeling like they’re already behind.

Below is a recap of the conversation, but you can watch the full webinar HERE

First things first: What actually makes something an AI agent?

Chatbot. Copilot. Assistant. Agent. We’ve accumulated a lot of names very quickly.

Jeet offered a useful distinction: an AI agent has a goal, can plan how to achieve it, and can connect to other tools to take action across multiple steps.

Dave made it even simpler.

Think of a chatbot like sitting in a conference room with a very smart coworker. You talk through the problem together, but everything stays in the room.

An agent is when you leave the room, hand that coworker your computer, and send them off to actually do the work.

That ability to move from answering to doing is where things get interesting.

AI has finally escaped the chat window

The biggest change over the past year isn't necessarily that AI got better at writing.

It's that it can now connect to the places where work actually happens.

Jeet shared how agents can pull from meeting notes, Google Drive, customer conversations, and other systems to surface insights and trigger workflows. At CandorIQ, Haris shared how agents can help process customer conversations, flag sentiment, create tickets, structure data, and surface the right information to the right person.

That means we're moving from:

“Summarize this meeting.”

to:

“Summarize this meeting, flag the important issues, update the right systems, create the next steps, and tell me what needs my attention.”

That's a much bigger shift than simply getting a better chatbot.

Your first AI agent doesn't need to be revolutionary

One of our favorite examples came from Dave.

He had been manually turning HRIS exports into Google Sheets with pivot tables, formatting, and reporting for a client.

So he gave his existing spreadsheet to AI, worked with it to create something better, and eventually turned the process into a reusable skill.

Now he can drop in a new HR export and get a significantly more sophisticated report in five to ten minutes.

No robot CHRO required.

Sometimes the best place to start with AI is simply asking:

What annoying thing do I keep doing over and over again?

Want your team to use AI? Give them permission to be bad at it first.

Buying everyone an AI license isn't an adoption strategy.

Dave boiled successful adoption down to three things:

Leadership support. Investment in tooling. Space to try and fail.

Jeet has seen hackathons become one of the strongest ways to build AI fluency because they give people permission to experiment without worrying whether they're “doing AI right.”

In one example, employees were initially hesitant to even share what they were building in an AI Slack channel. After a hackathon, they were sharing so much that the channel had to be muted.

The lesson?

You probably don't need another AI training deck.

You need people actually using it.

Don't ask AI to make the decision. Ask it to challenge yours.

This might be one of the easiest AI habits to steal from the conversation.

When using AI for decision support, Dave doesn't necessarily start with:

“Tell me what to do.”

He starts with:

“Here's what I'm thinking. Critique it.”

That changes AI from an authority into a thinking partner.

Jeet agreed: AI can pressure-test a framework, surface missing information, and make analysis faster, but the final decision still belongs with a human.

Especially when those decisions affect someone’s pay, role, career, or employment.

In workforce decisions, AI needs to show its work

If AI recommends a compensation adjustment or surfaces a workforce insight, “trust me” isn't good enough.

Jeet discussed the importance of grounding AI in company-approved information and pointing users back to the sources behind an answer.

Haris shared a similar philosophy at CandorIQ: AI shouldn't be a black box. If it reaches an answer, leaders should be able to understand the data and framework behind it.

Think back to school:

The answer matters. But so does showing your work.

That becomes even more important as agents move from administrative workflows into actual decision support.

Plot twist: AI might not make us work less

AI is usually sold around one promise: save time.

But Dave shared that something slightly different is happening.

When AI saves him time on repetitive work, he often puts that time straight back into improving the quality of the output. Instead of asking AI to make one deck faster, he can create five approaches and decide which one is best.

Haris and Jeet echoed the same experience: less time on administrative work doesn't necessarily mean less work.

It can mean more time for strategy, relationships, judgment, and higher-quality decisions.

So perhaps the AI productivity story isn't simply about doing the same work faster.

It's about being able to do work that previously wasn't practical to do at all.

Expertise isn't becoming less important

AI can give a very convincing wrong answer.

That's why Dave still puts a lot of value on expertise: when you deeply understand a subject, you know when something doesn't look right and where to push back.

Jeet sees this changing the shape of people's skills.

Instead of being T-shaped—broadly capable with deep expertise in one area—AI could help more people become M-shaped, developing depth across several areas while AI helps them learn and execute faster.

But judgment, empathy, creativity, and actual subject-matter expertise still matter.

Possibly more than ever.

Final Takeaway

Dave summed up the current AI moment particularly well:

“People are further ahead than they feel, but further behind than they think.”

LinkedIn can make it seem like everyone else has already built an army of AI agents running their company while you’re still figuring out your prompts.

They haven't.

At the same time, many of us are probably using only a fraction of what the tools we already have can do.

So you don't need to transform your entire company tomorrow.

Pick one workflow. Experiment. Give your team room to fail. Keep humans involved where judgment matters. Then turn what works into something repeatable.

Or, in Jeet's much more memorable words:

“Burn some tokens. Ignore the noise, and give it a go.”

Watch the entire webinar HERE

‍