OpenAI’s Dollar-Menu Option
I’m not using AI to write this post.
To be clear, I use AI *all the time* in my professional life. It’s improbable at this point to think anyone doesn’t use it. In the past two years, the technology has enjoyed exponential adoption in the private sector, even when people claim they don’t use it. I’m not ashamed of it. AI has accelerated my ability to take on new projects with increased output. And while some people may think that’s a lazy approach, let me remind you: I’m a consultant. I say things like, “Well, [so-and-so customer], it’s not the cost of doing business that’ll kill you, it’s the cost of not doing business.” Coming up with gold like that all day isn’t as easy as it sounds. Of course I’m going to use AI.
All joking aside, today I’ve decided to write this post without any AI assistance. I read an article the other day about how AI is potentially robbing us of our ability to think, just like social media is potentially killing our ability to be conscientious. I think both claims have some merit, mostly because I don’t have the time or intellectual chops to argue against scientists from MIT or the editors of the Financial Times. But I also like to think I use AI in a manner that acts more as an accelerant, not a replacement, for creative thinking. It isn’t exactly Skynet, but I do understand the inherent risk in being overly dependent on AI’s capabilities.
There’s a lot of discussion about the “dangers” of new technology anytime humanity makes a discovery or leap in a particular field. The classic risk/reward paradigm persists. Our preternatural need to touch the fire before we figure out it burns is unavoidable. (My wife tells a funny story from the early aughts about her then-preteen cousin, who, within minutes of discovering Google, was caught searching for “hot grills” by his dad.) Simply put, we are in the exploration phase of AI, so we are all going to read a lot of articles about the “what if” of it all.
Which is why I’m just a little concerned about the news that OpenAI is giving ChatGPT Enterprise to the Federal government just a dollar a year per user. I spent 11 years in Federal government trying to convince any Chief Information Officer (CIO) who would listen to adopt new technology, especially software as a service (SaaS) products, to drive efficiency and productivity. I was there when Salesforce and ServiceNow got FedRAMP’d, before the SaaS goldrush on the public sector. I was repeatedly frustrated in my efforts to unlock the CIO whitelist of available software, but I also understood why those lists existed.
At the end of the day, when all the cables are pulled and the servers are racked, federal CIOs exist to serve as the indelible high wall of compliance at the end of the grueling death march that is the software procurement process. I remember getting a demo of Smartsheet and being told by my CIO at the time that “no one will ever trade Microsoft Project for that security nightmare”. Oops. I even awarded a contract to one of the big four consulting firms, and part of their proposal was designing and building a custom python-based dashboard for enterprise access across that agency to report, among other things, return on IT investments. The CIO scratched that section out in his review of the statement of work, called me into his office and with a straight face, asked me, “Are you trying to get me fired?”
I guess this is a little bit of me having my cake and eating it, too. I admit, it’s confusing. I’ve talked here about public sector AI adoption and lamented how slow that appears to be going. And now I’m complaining about the potential for AI adoption to be accelerated at the Federal level in an unprecedented manner. Call my cynical. For starters, it appears GSA has sidestepped a competitive process for procurement. And for all its altruistic window dressing, this could just be step one in OpenAI’s master plan to take over the public sector by capitalizing on the millions of procurement-related prompts that will be entered in the first year of federal adoption. (Artificial intelligence? More like market intelligence, am I right?)
I am confident most Federal employees have used ChatGPT in their personal lives. I am also confident that some of them have used it professionally, even with the restrictions on using personal devices for any official business in government. I am not confident, however, that the CIOs overseeing implementation of ChatGPT enterprise have any interest in training agency workforces on how to adopt AI into their various workflows.
In 2023, I had a federal client make an early attempt at folding ChatGPT Enterprise into their agency intranet. It was deployed with no training or education for the user. As a consultant, my pleas to deploy an agency-wide desk guide to maximize return on what was then a significant investment (more than a dollar, to be sure) were ignored. This agency simply wanted to be first to board the AI train. Well, that’s fine, but this AI train happened to have no seating inside and our first destination was 18 months away, when the first feature milestones were planned. Unsatisfying doesn’t begin to describe the outcome.
That all may seem pedantic, but for me, it really is a perfect storm of government waste and private sector wall-eye: the economic and climate cost to run large language models (LLMs) is well-documented. We’ve just deployed the most popular of these LLMs to the entire Federal workforce for a dollar in the first year (the price goes up after 12 months). And there’s no perceivable plan to adopt and improve the tool over time. Oh, and the ChatGPT user experience keeps getting objectively worse. No, my biggest problem isn’t how or why this is happening so much as it is that it’s happening now, at a moment when AI could use a shot of good old-fashioned, government incented competition.
Some of the biggest leaps in technology were the result of the government simply asking for them. For the second half of the 20th century, we experienced a wholesale renaissance in military defense and space exploration thanks to the federal government. If AI truly signifies the next horizon of a technological manifest destiny, why relegate the market leader to a dollar-store version of its true potential? Imagine a $1 trillion indefinite delivery, indefinite quantity vehicle for any qualifying AI to compete across the government for their business leading to yet another awakening in technology evolution. It’s also important to remember we are just a short decade out from quantum computing being widely commercially available, and AI is just one tool coming into focus that will serve as a necessary precursor to technologies we have yet to conceive. In other words, instead of treating AI like it’s the last in a series of advances, we need to start treating it like it’s the first and make the investments necessary to improve it. Otherwise, all the government did was buy a discount ticket to board the aforementioned AI train, once again trading true innovation for the predictable unrelenting embrace of the status quo.
In all honesty, I was going to end this post right there on an admittedly bleak point. But here’s the good news. Maybe this investment, however limited, creates a wave of excitement for AI in the public sector. Maybe that enthusiasm trickles down to state and local, and OpenAI is able to democratize access to its enterprise tools across a multitude of communities, regardless of economics. Believe me, I want that to be the case. In fact, if I had the opportunity, I’d tell Sam Altman he shouldn’t stop with federal. He should be giving it away for a dollar to every public sector employee in the country. In fact, after a couple days of thinking about this subject, it now seems strange to me OpenAI stopped short of a nationwide public sector bid. Either OpenAI considered this option and concluded that on a state-by-state basis, the procurement challenges are just hard to get around, or they completely ignored the option because their eventual goals didn’t align with the whole Oprah “everybody gets a car” model. Whichever the case, it’ll be interesting to see how ChatGPT’s adoption plays out in the federal space.
Of course, the real test isn’t what happens in GSA’s press releases or during the first annual license review. It’s how government evolves when new tools become mundane, not exceptional. We should be looking for evidence that agencies will craft use cases beyond procurement form templates and basic email drafts. I want to see federal teams harness AI for programs that truly serve the public so we stop fixating on cost and start asking what’s possible. Things like climate modeling, emergency management, and healthcare.
That sort of change requires leadership, training, and a willingness to iterate in public, not just behind closed doors. If this turns into another example of “check the box, move on,” the public won’t see returns greater than a dollar a head. But if this is the foundation for an era of a new digital-first government, we could look back in a few years and say that it took ordering off the “AI dollar menu” to figure out we wanted something better.