Mark Johnson

Helping Federal orgs get more from emerging tech. Husband, father, retired Navy pilot. Databricks employee, but all opinions all mine and only mine.

[Mark Johnson @MarkJohnsonIT]

Every AI demo happens in a cleaner environment than the average enterprise. That's why deployment is so much harder than the demo. Models are improving quickly. Untangling years of systems, processes, and data takes much longer.
https://t.co/JjVcq3YUrY

[Mark Johnson @MarkJohnsonIT]

The recent breaches committed by AI models from OpenAI and Anthropic demonstrate why common standards and testing programs are important. A neutral, trusted body needs to be in place to oversee these models.
https://t.co/2RnyGxEDdP

[Mark Johnson @MarkJohnsonIT]

One thing I respect: admitting reality. Zuck said Meta's development using AI agents has been slower than expected. That's refreshing. Companies learn faster when leaders are honest about what's working, what's not, and what needs to change.
https://t.co/s1f67iShqI

[Mark Johnson @MarkJohnsonIT]

I've always told federal clients: I can fix a lot of technology problems. I can't fix your policy problems. AI is exposing how many policies were written for a different era. If we want AI to deliver actual value, the policies need to evolve alongside th… https://t.co/cJGU7h5DG6

[Mark Johnson @MarkJohnsonIT]

AI deployment fails when we ignore the human element. One tip for federal leaders: Fix the "role friction."
Clearly define where the machine's work ends and human authority begins.
#GovTech
https://t.co/nlKUP1WNnO

[Mark Johnson @MarkJohnsonIT]

AI agents create a new management skill. Most leaders know how to hire, coach, & organize. Far fewer have experience deciding what work belongs with a person, what belongs with an AI agent, and how those two work together. Figuring it out first will give you an advantage that's…

[Mark Johnson @MarkJohnsonIT]

AI adoption isn't a tech problem; it's a people problem. Two ways federal IT leaders can win:
1. Map employee readiness archetypes.
2. Build trust, not just tools.
#AIGov
https://t.co/yQlgyje8gy

[Mark Johnson @MarkJohnsonIT]

During rapid change, teams rarely run out of ideas—they lose alignment. Before discussing tactics, reconnect on purpose: What are we trying to achieve? What hasn't changed? What tradeoffs matter? Clear answers create focus, and focused teams execute faster.

[Mark Johnson @MarkJohnsonIT]

91% of respondents said they're creating new AI budgets rather than cutting existing software spend. If that trend holds, AI isn't following the adoption pattern most enterprise software did. That suggests we're still early.
https://t.co/MeTqFw8wQg

[Mark Johnson @MarkJohnsonIT]

I think data management will be one of the defining leadership challenges of the AI era. Two teams can look at the same business & produce different numbers because they pull from different sources. When those inconsistencies feed AI systems that people rely on for major decisi…

[Mark Johnson @MarkJohnsonIT]

Watching Kellie's organizing work has reinforced something a core belief: people underestimate the value of structure.

Disorganization creates friction. Time gets spent looking for things, revisiting decisions, and navigating confusion. Good systems free people to focus on wha…

[Mark Johnson @MarkJohnsonIT]

A few years ago, it would have been strange for an AI conversation to focus on electricity. Today, data centers are buying power developers as they struggle to meet demand.

I didn't expect one of the more interesting AI stories of 2026 to involve power companies, but here we a…

[Mark Johnson @MarkJohnsonIT]

On my way to #DataAISummit in SF! If you're here, lmk and let's talk federal IT solutions!

[Mark Johnson @MarkJohnsonIT]

The DoD's AI Acceleration Strategy is directionally right and the $30B budget request for infrastructure is also good. Now let's work together to show how to securely share data so everyone who needs it, has it!
https://t.co/8kQoBmWWcS

[Mark Johnson @MarkJohnsonIT]

DoD's contested logistics sprint includes an AI tool to help commanders make real-time decisions. The hard part: connecting data across commands. AI is as good as it's data.

https://t.co/tx9FKvPJb8

[Mark Johnson @MarkJohnsonIT]

$29.5 billion to centralize & scale DoD's AI computing infrastructure is a huge investment. I'll be interested in seeing what gets ingested into all that compute. $30 billion of infrastructure running on poorly governed data might produce faster decision… https://t.co/lSlhc76XGS

[Mark Johnson @MarkJohnsonIT]

A miscoded field in one program office is a minor annoyance. The same field ingested into an enterprise AI system across twelve offices informs four months of award decisions before anyone notices. Scale only buries data problems deeper and spreads them wider.

[Mark Johnson @MarkJohnsonIT]

894 million tokens per day in agentic workflows during Operation Epic Fury. Compute bottleneck is a serious concern. But I'd push back gently on where the focus should actually be. The data governance layer underneath those workflows is the constraint ge… https://t.co/3rmixIqZn2

[Mark Johnson @MarkJohnsonIT]

Stop by the @databricksinc.bsky.social table at the AFCEA Bethesda Health IT Summit this week and let's talk about enabling the workforce with better data and analytics to improve health outcomes!

[Mark Johnson @MarkJohnsonIT]

Eight AI vendors now have agreements to deploy on DoD's classified networks. Buried in the announcement: it's not clear how the department plans to use them. Access and readiness to deploy responsibly are different problems. Adding vendors doesn't solve … https://t.co/JUV4J8hcUv