Ashish Bhutiani

Hopeless sports fan. I do things @splunk.

[Ashish Bhutiani @ashishterp]

There's an interesting irony in AI and Customer Success.

The more we automate the work around the customer, the more time we can create with the customer.

Agentic workflows can absorb more of the operational load and give CS teams room to get back to understanding the customer's business.

To me, this feels like a much better use of AI.

[Ashish Bhutiani @ashishterp]

Building a business from scratch teaches you fast how connected every decision is.

I've felt this running the businesses I own outside my corporate career: a pricing change that shifted demand, an order that left a customer without what they needed for an event, a supplier switch that added options but slowed delivery.

Every time, I learned the same lesson: the day-to-day decides more than the plan does.

You can have a strong concept. Sustainability comes down to the daily choices around operations and where you spend your time, and those choices compound fast when you're the one who owns the outcome.

Vishal Vivek makes a similar point in his piece on building sustainable businesses: you earn credibility in the field long before anyone sees it.

It's changed how I think about growth too. Sustainable growth means you can keep delivering well as demand grows. The discipline behind that matters as much as the original idea.

What's a small decision in your business that ended up mattering more than you expected? https://t.co/JIixrq82SY

[Ashish Bhutiani @ashishterp]

I paid for organizational knowledge once. It's called a franchise fee.

The whole model is buying someone else's operating knowledge: the playbook, the vendors, the layout, the mistakes already made. And the strange part is that with all of that identical, units still perform differently. The difference is what the operator knows about their own market and their own people (what rain does to a Saturday, which local events actually move traffic, which marketing works here but not two towns over). The playbook can't hold that.

Enterprise AI works the same way. Every company can buy the same models now, but what they can't buy is knowing how their own work needs to move.

If you're betting on AI for advantage: what do you know about your operation that isn't written down anywhere?

[Ashish Bhutiani @ashishterp]

Jeffrey Saviano's argument in MIT Sloan is that leaders have to decide what their organization stands for on AI. Fair enough.

In agentic systems, though, somebody already drew that line, and it probably wasn't a leader. It was whoever granted the service account its permissions.

That was an ethics decision, but the person making it was closing a ticket.

I work in identity security, and the gap I see is a board debating oversight in one room while an engineer in another gives an agent write access so a workflow stops failing.

If you're drafting AI governance policy: has anyone mapped it against the permissions your agents already hold? https://t.co/RdC2MJDaRW

[Ashish Bhutiani @ashishterp]

AI governance gets real when you ask one question: six months from now, could you reconstruct what your agent did today?

Most teams can answer that for a user. Fewer can answer it for an agent that called four services, borrowed a token somewhere in the middle, and changed a record on the way through.

That's an identity problem before it's a governance problem. If you can't say which identity the agent acted as at each step, you don't have a log. You have a story.

If you're piloting agents: could your team answer that for last Tuesday?

[Ashish Bhutiani @ashishterp]

Vivek Ahuja has sat through dozens of enterprise AI evaluation meetings over the past two years. He says most open with a slide comparing GPT-4, Claude and Gemini benchmark scores.

His argument in Forbes: the number worth trusting is the one your own team measures in production, against your own edge cases. He puts the gap between what vendors report and what his teams see live at 15 to 25 points.

If you’re running a deployment right now, who measured your accuracy? You, or the vendor? https://t.co/2wDIuoWAja

[Ashish Bhutiani @ashishterp]

Cloud migrations have a way of exposing how a business really works.

The infrastructure may initiate the move, but the bigger opportunity is preparing the organization for what changes around it.

Treating change management as part of the architecture early is what turns cloud flexibility into lasting business value.

[Ashish Bhutiani @ashishterp]

The clearest sign a company needs Customer Success isn't in the customer count. It's in the Professional Services backlog.

When PS starts getting asked for things that were never in the statement of work, like a workflow question, a "can you just show us how", that's not a services problem. Customers are asking someone to own the outcome and PS is the only warm body available.

I lead both functions globally, and I'd say that's the most reliable early indicator I see. It shows up months before retention does.

If you run services or support: how much of your team's week goes to work that isn't in any contract?

[Ashish Bhutiani @ashishterp]

Running a boba shop and an ice cream truck has been a good reminder that the fundamentals do not change.

You are still working with vendors, building partnerships and aligning value on both sides. The scale looks different, but the thinking stays the same.

When both sides understand what they are trying to achieve, things start to move faster and with more purpose.

[Ashish Bhutiani @ashishterp]

The subscription is almost never the expensive part.

Kevin Smith's CFO Dive piece makes this the second of five questions CFOs should ask before an AI bet: do you actually know the true cost? The license is visible. Data prep, integration, training, governance, and the people who babysit it afterward usually aren't.

I get to see this from two seats. One is a global team at Palo Alto Networks. The other is a franchise I own, where the AI bill itself is small, but the real cost was change management and training.

If you rolled something out this year, what was the cost that never made it into the business case? https://t.co/AZlowQ9B2T

[Ashish Bhutiani @ashishterp]

Loyalty looks different in enterprise vs consumer, but it still has to be earned.

In enterprise, people may not choose the software, but they decide every day how much they engage with it. That shows up in adoption and whether it becomes part of how work actually gets done.

When that connection is there, the product starts to feel natural. When it is not, even strong functionality can feel like friction.

[Ashish Bhutiani @ashishterp]

3% of leaders say they feel highly prepared to lead AI-enabled teams. I read that and mostly felt relief.

I lead a global team in identity security. We're rethinking how we deliver with AI right now, and honestly, I wouldn't put myself in the 3% either.

I've stopped expecting to feel prepared. That word assumes the path already exists and you studied it. This one is forming while the decisions come due.

So I aim for steady instead. Make the next call and say what it's based on. When it misses, adjust in the open.

What does "prepared" even look like in your world right now? https://t.co/TVNrwJ3Psl

[Ashish Bhutiani @ashishterp]

Look where full automation is already normal in this chart: routine cognitive tasks. The predictable stuff. Everywhere else, people keep a hand on the wheel.

The common assumption is that automation expands naturally as trust grows. I don't buy the "naturally." Trust gets built deliberately: when a team can see what the AI did, check it cheaply, and undo it when it misses.

My bet: the tasks that go fully automated first won't be the easiest ones. They'll be the most checkable ones.

If that's right, the fastest path to the agentic future isn't better models. It's better plumbing.

What made your team comfortable handing over its first full workflow?

[Ashish Bhutiani @ashishterp]

Every week there's a new model, a new tool, some post declaring that everything just changed (again). And every week I watch smart teams consider changing direction because of it.

The leaders who handle this well aren't the fastest adopters, but they know exactly what they're optimizing for, so most announcements just get a shrug. Not in a dismissive way. The kind of shrug that comes from knowing this news won't get them to their goal any quicker.

They still read everything, by the way. They just don't confuse staying informed with being obligated to act.

Saying yes to everything feels like momentum. A lot of the time it's just drift.

What's the last AI announcement your team decided to ignore?

[Ashish Bhutiani @ashishterp]

92% of finance leaders say they're under pressure to prove AI ROI. Only 22% of organizations say returns have met expectations.

Different surveys, weeks apart, describing the same squeeze: nearly everyone has to show value, hardly anyone is sure they have it.

Chris Dimitriadis's piece below argues the gap is governance, and I think he's right, with a little more context. In practice it comes down to two questions most orgs can't answer: who approved this use case against which metric, and who has the authority to kill it when the metric doesn't show up.

Teams that can answer both easily have a much better chance of living in the 22%.

Which one can your org not answer yet? https://t.co/wyVmyetP56

[Ashish Bhutiani @ashishterp]

92% of finance leaders say they're under pressure to prove AI ROI. Only 28% require audit logs of what their AI agents actually do.

Read those together and the risk is obvious. The pressure to show returns, without the plumbing to measure them, produces numbers that look good in a deck and mean very little in reality.

I'd still take the pressure over the alternative. Two years of "experimentation" with no finance questions is how AI programs die quietly. But the order matters: instrument first, then promise.

If your CFO asked you which AI initiative is paying for itself, how confident would your answer be? https://t.co/SNsgBCWCO5

[Ashish Bhutiani @ashishterp]

When your global team runs smoothly, is it because everyone adapted, or because one culture quietly won?

Most global teams develop norms without ever choosing them. They tend to drift toward whoever holds the most authority. Their communication style, their pace, their definition of what good looks like. Nobody plans it that way.

I've led teams across geographies long enough to know the friction almost never comes from etiquette. It comes from unspoken assumptions about how trust gets built, how decisions get made, and whose way of working gradually becomes the one everyone else adapts to.

The teams that navigate this well tend to do something early on. They talk about it. "How do we want to work together?" "What do we each bring that's worth keeping?" It sounds basic. It changes more than you'd expect. https://t.co/dcJuhLNA5E

[Ashish Bhutiani @ashishterp]

Every team deploying AI agents eventually hits the same question: how much access should this thing get?

Give it too little and the agent is useless. Give it too much and you've created your most privileged employee, one that could make mistakes at machine speed and never stop to ask whether something feels off.

No model should define your risk tolerance. A person draws that line, then redraws it when the environment shifts.

Curious where other identity and security folks are drawing it.

[Ashish Bhutiani @ashishterp]

An AI agent burns 5 to 30 times more tokens per task than a chatbot, per Gartner. That traffic crosses networks most AI business cases still book as a fixed cost.

Pilots keep networking cheap because they stay in one region with predictable traffic. Production workloads move data across clouds and call APIs on patterns the capacity plan never modeled. Procurement cycles lag AI adoption by 12 to 24 months, so teams discover the cost after deployment instead of budgeting it up front.

When EMA surveyed 269 IT professionals on their biggest networking challenge for AI, security beat budget, 39% to 34%. I see the same pairing inside migration programs: each agent you deploy adds traffic on the wire and credentials someone has to own. https://t.co/2w3SDvUOpg

[Ashish Bhutiani @ashishterp]

The brands scaling fastest right now have figured out that the product itself is the marketing. People post the drink before the straw goes in. That changes how you think about menu design and even which markets you enter first.

The article's point about brand recognition traveling ahead of you matched what we saw at opening. Chatime had fans in Las Vegas before we ever turned on the lights, people who knew the brand from other cities and other countries and walked in already sold. That demand existed before we spent a marketing dollar.

The other claim is operational simplicity: the easier a concept runs consistently, the faster it scales, and complexity is what breaks franchise models somewhere around the second or third location. I can't confirm that from experience yet. It's the thing I'm watching most closely as we get past opening.

So those are my big indicators for the next year. Does the product keep marketing itself once the novelty fades, and does the operation stay simple as we grow. https://t.co/JDO13J44Kd