Your Workplace Ops Runs on Spreadsheets and Slack. AI Is Coming for That (In a Good Way)

Think about how workspace access, approvals, budgets, and policies actually get managed at your company. Odds are it looks something like this: a spreadsheet somebody owns, a Slack thread where somebody asks permission, a calendar invite, a manual approval, and three tools that don't talk to each other.

That's not a tooling gap. That's the operating system of most workplace teams in 2026.

In a live conversation on the Croissant channel, Zoltan Szalas, CEO of Croissant, sat down with Matt Roberts, co-founder of Happy Operators, who spends his days helping companies build with automation and AI. The question on the table: how does AI actually change the manual, duct-taped layer of workplace operations? Not in theory. In practice, for People, Ops, Workplace, Facilities, and Chief of Staff leaders who have to make it work.

Here's what came out of it.

Your employees are already using AI. The only question is where.

Matt's first point cuts through the "should we adopt AI?" debate entirely: your team already did.

"Employees are doing it anyway without you knowing," Matt says. Shadow IT is real, and AI made it effortless. You have two options: disincentivize it, or accept that the technology is driving real business value and give people the ability to use it inside your ecosystem instead of outside it.

That second path requires leadership to accept some risk. But the alternative is worse: your workplace data, your policies, and your processes flowing through personal accounts you can't see.

And the goal isn't headcount reduction. "It's not about trying to fire people," Matt says. It's about moving humans to strategy, creativity, and better ways of working together, and moving admin to the machines.

Where do you actually start?

Not with a transformation program. With a screenshot.

Matt's favorite starting move is almost embarrassingly simple: take a screenshot of your calendar, hand it to a frontier model, and ask "what on here do you think we could apply AI to?" Then flip the dynamic: get the AI to ask you questions instead of expecting yourself to have the answers.

The skill that matters now isn't prompt engineering. It's context. Matt's rule: treat AI like a new hire, not a magic box. You'd never tell an intern "just start your job now." You'd explain what the company does, what the team does, and the three or four things you check to get a task done. Talk to AI the same way. And don't feed it a broken process and expect a fixed one back. Be honest about what works and what doesn't.

How do you roll it out across an organization?

For the HR leaders and chiefs of staff in the audience, Matt laid out the playbook he uses on real client projects:

1. Get leadership in a room first. CEO, COO, HR director. Not to approve a budget, but to talk honestly about how work actually gets done today. Run a transcriber in the background so nobody has to take notes. The transcript becomes your raw material.

2. Find the small wins before the big ones. "You need to provide value early on," Matt says. "If it doesn't work two times, people get frustrated and turned off by the technology." A big win like automating project estimation sounds great, but it touches a dozen spreadsheets and can go wrong in a dozen ways. Bank the small wins first. Sometimes the win is just showing an HR director features that already exist in the tools she pays for.

3. Expose people through multiple channels. One-on-ones, workshops, hackathons. And critically: give people time on the company's watch. Three hours where nobody has client work and everyone builds something internal beats any mandate to "start using AI."

4. Shadow, don't shove. Senior people won't offload their work to an agent on day one, and they shouldn't. Matt's technique: sit alongside them while they do the task their way, then add AI to the process one tweak at a time. Nobody has to change their behavior from the word go.

Should everyone get access to the company brain?

Zoltan raised the setup Croissant runs internally: an operational brain with the CRM, engineering tickets, sales-call transcripts, and enrichment tools all connected, so the AI can connect dots across the company. His question: what happens when you give that to every employee?

Matt's answer starts with a question back: what's your risk profile? If your connected tools authenticate against each person's own account, employees don't see any data they couldn't already see. What they gain is leverage. "The closer to the problem you are, the more chance of succeeding in trying to solve it," Matt says. Give the person who owns the problem the ability to work through it at scale, and problems get solved much faster.

Then he shared the twist worth stealing: make the AI public. Shopify deployed an internal agent with one rule: nobody could talk to it privately, only in public channels. Matt runs the same experiment with a client agency, where the shared agent (named Steve) lives where everyone can watch it work. When AI conversations happen in Slack or Teams instead of a private window, everyone learns from everyone's prompts, the experts correct the AI in the open, and adoption compounds instead of fragmenting.

What about the fear?

Zoltan brought up a conversation with a 100,000-person company hesitant to roll out AI because of the anxiety it creates. The tension is real: innovation on one side, employee wellbeing on the other.

Matt's answer: show, not tell. Leaders who demonstrate AI handling their own drudgery ("I only check email three times a day now, and here's why I can have this coffee with you") land differently than leaders who mandate a tool. And frame the shift as a company problem, not an individual threat: we all have to re-engineer our roles, we're doing it together, and you upskill on company time, not your weekend.

Both agreed on where this lands for careers. The employees who thrive will be outcome-based, not task-based. As Zoltan puts it: if AI does the research, digs through the transcript, and handles the little things in between, your job becomes the outcome itself. Close the deal. Make the workplace work. Matt sees whole roles reshaping around this already: designers shipping working prototypes instead of Figma files, salespeople becoming go-to-market engineers who wire tools together.

Their 2027 predictions? Matt says everyone becomes a builder. Zoltan says regular AI use becomes a benchmark skill, like email: not using it will be a deal breaker.

The easiest win, and the bigger point

Asked for the single easiest workplace process to automate first, Matt didn't hesitate: transcribe your calls. Then route the transcript wherever it's useful: the CRM, the project tracker, a Slack summary for everyone who wasn't in the room. It's one click, and for distributed teams it quietly replaces an entire category of "did anyone write that down?" overhead.

That answer is small, but it points at the bigger truth of this conversation. The problem in workplace ops was never a missing feature. It's that the connective tissue (who gets access to what space, who approved it, what it costs, whether the policy was followed) lives in spreadsheets and chat threads. The spreadsheet is the real competitor, and AI plus automated workflows is how it finally gets retired.

That's exactly the layer Croissant works on. Croissant helps companies manage distributed workforces with governed access to flexible workspaces: one place for workspace access, policy, budgets, and usage visibility, instead of a stack of manual workarounds. If your team's workspace operations still live in a spreadsheet and a Slack channel, start at getcroissant.com.

Watch the full conversation:Workplace Ops Is Still Run on Spreadsheets and Slack. Here's How AI Changes That (with Matt Roberts)