The Shelf Life of Enthusiasm
And how to preserve it!
AI makes noise. Big conferences. Jaw-dropping demos. They are all classy. Stylish. Sometimes overwhelming?
The enthusiasm is real. But it may not translate into tangible organisational output. Why? Because enthusiasm has a shelf life.
If you don’t institutionalise it, it expires. Fast.
At DeepIntent, we knew this. So instead of chasing the hype, we designed an operating model to absorb it. Not just a dashboard. A way of working.
Remember, we’re the PMO team - very tech but traditionally non-AI. But over the weeks following a (very) successful hackathon, we knew that we had to embed AI into our operations. We didn’t need more tutorials. We needed rituals, systems, and muscle memory to harness AI.
This article isn’t a “look what we did.” It’s an implementation playbook for anyone trying to capture and institutionalise AI hype into long-term momentum.
Understand the Hype
AI is the new electricity… or fire, depending on who you ask. The point is, everyone has a take, and all of them are right.
We didn’t dismiss the hype, but we didn’t want to get consumed by it either. We observed. We watched tutorials on how people were actually using it. What worked? What didn’t? We understood the narrative before stepping into the arena.
Now, we’re a team in a sophisticated tech startup, so it goes without saying that we’re already tech-forward. So we never had “fully manual” workflows as such. We used every tool you’d expect a PMO team to use in a modern startup. But what we didn’t use in our day-to-day was AI.
Yes, we used ChatGPT and other usual suspects, but nothing that allowed us to truly harness the power of AI to multiply our output.
Then came the much-awaited DeepIntent AI-Hackathon. A focused, intense, three-day, in-office sprint. Being in the middle of the action helped all of us finally grasp the lingo and more importantly, the application.
What’s an LLM? What are tokens? Why are there so many models? Is one better than the other? How do you know?
The hackathon became our launchpad into AI orbit.
Everything changed after that.
Unlock Time to Learn It
Any orbital change is tough. So was this.
And here’s the honest truth: learning AI takes time. Hearing podcasts on your commute won’t cut it, and watching a demo isn’t the same as understanding how it fits into your workflow.
Now that we understood the vocabulary and had a foundational understanding of AI and its applications, it was time to remove the training wheels and actually start using it.
But that required something most teams struggle to find: time.
The question on the table was: Now that we’re knowledge-rich, how do we become time-rich to apply this knowledge?
That’s when Sourabh Gandhe decided to launch an all-out attack on meetings!
This is the part I absolutely love about startups. Changing cadences in any large company would take weeks, if not months. Also, a meeting would likely be held to decide whether we should reduce meetings.
But not at DeepIntent.
It took us a day to evaluate our meeting framework, and another day to change it. We don’t work for processes. We make them work for us. We’re not prisoners to manifestos or implementation guides. This, my friends, is the true definition of agility. Anyone telling you otherwise is lying. This shift in our meeting framework unlocked hours of focused time across all roles and hierarchies.
We used that time to explore AI with a clutter-free mind. We played with prompts. We watched YouTube videos. We asked basic questions. We broke things. We wrote bad scripts, and then slightly better ones. Slowly and steadily, we built our muscle memory.
We now had both the knowledge of AI and the time to use it.
Leverage Internal Expertise to Know the Guardrails
Every company has a few people who are ahead of the curve. They’re the ones quietly writing those Python scripts or figuring out how to plug LLMs into their workflows. We found those people, sat with them, and learned not just what we could do with these tools but, more importantly, what we shouldn’t.
This wasn’t about becoming AI experts. It was about understanding the boundaries. Don’t forget we work in a healthtech company. Data privacy is sacred and non-negotiable.
LLMs can hallucinate. API access comes at a cost. Compliance isn’t optional. These things don’t show up as you start playing around, but they will blow up in your face if you have not taken the time to understand the guardrails.
AI isn’t plug-and-play. It’s plug-and-think. So we thought before we played.
This helped us build with intention, not just excitement.
The next question then was: Where exactly does AI fit into the PMO world?
Evaluate the Fit into Your Function
This is the fork in the road. Some workflows needed to be automated - routine reporting, data pulls, sprint summaries. Others needed to be made intelligent - insight generation, retrospective analysis, anything that can support our decision-making.
We didn’t treat AI like a sledgehammer. We used it like a scalpel - carefully, precisely, always asking: Does this make our process faster, smarter, or noisier?
AI experimentation isn’t a single-variable problem. You’re not just trying to get something to “work.” You’re balancing time, cost, governance, and clarity simultaneously.
We started by identifying operations that scale linearly with team size. These are the best candidates for automation because the cost-benefit math becomes obvious. As the organisation grows, manual effort scales with it …. unless you intervene. We did.
We mapped recurring workflows - some daily, some weekly, some bi-weekly and calculated how much time they actually consumed.
The answer? Over 25 days of work per month. Not kidding. 25 days!
That's the equivalent of an entire full-time role and change!
We had a job to do, and we didn’t want to hand this off to a tech team. We wanted to build it ourselves. That’s the underrated move: giving AI tools to a team that knows the function deeply.
As of this writing, the PMO team, mostly from non-tech backgrounds, has written App Scripts to automate recurring reporting, Python scripts to trigger routine updates and communications, and even reimagined Slack announcements based on system triggers.
While doing so, we’ve kept one principle front and centre: don’t let perfect be the enemy of good.
And it’s working. We've already saved hours, not just for the PMO team, but for the broader organisation.
We were able to get—and give—the gift of time.
And This Gift of Time is indistinguishable from Magic
Once the dust settled, we weren’t working less; we were working better. Less time running those standups and sharing MoMs. More time coaching on how to enable agility.
AI didn’t take our jobs. It took the junk out of them.
There are a million things waiting to get done in a startup. Overflowing backlogs, packed roadmaps. And somewhere in all of this are your humble JIRA tickets that have the potential to become revenue-generating products.
But to even notice them, you need space to think. That’s what the gift of time gave everyone at DeepIntent.
Now that we’d freed ourselves from repetitive, manual work, we could finally think clearly about how to scale our function’s impact, without scaling headcount. That’s the unlock.
Time creates clarity. Clarity drives leverage. And leverage, not effort, is what scales teams.
I titled this article The Shelf Life of Enthusiasm because that’s what most hype ends up having - a short window before it fades.
This AI hype has been both overwhelming and exciting.
Overwhelming, because the pace can be anxiety-inducing. You feel the pressure to keep up, to react fast, to not miss the moment.
Exciting, because you know deep in your bones that this time it’s different. This isn’t just hype. This is the kind that actually changes how we live and work.
Glad to say we didn’t let it fade. We embraced, built around, and extended its life within DeepIntent.
We didn’t just ride the AI wave - we institutionalised it, one script, one prompt, one automation at a time.