AI for Government Services
Make GovTech great again
Hi! It’s been awhile. It turns out that maintaining a writing habit is hard. I’m trying to get back into it again.
If you haven’t already subscribed, please do so below!
Here are a few of the other posts I’ve written since my hiatus from Substack:
Now onto the actual post!
Interacting with the government today = Pulling teeth
How often have you heard someone say something positive about their experience dealing with the government? Probably never.
Yet social security, disability, health insurance, permits, visas, etc. are critical services that our government provides that we, as taxpayers, pay for. Once received these services can be life changing but there is a massive bottleneck at the administrative layer.
Disability claims, for example, take over 220 days to process which is 100 days longer than in 2019. This is 220 days of not being able to access funds that could potentially pay for treatment or supplement lost income. Every day of inefficiency here matters.
In some cases, the fuse is even shorter. For a 911 call center, every second of inefficiency matters. It could be life or death.
Often when we, as citizens, don’t receive these services in a timely manner we get frustrated at the people on the other end. We take our anger out on the poor rank and file worker of the Social Security Administration. While your mileage may vary, this is generally misguided! Often, the issue is the systems and process not the people. And that, very simply, means it can be solved by technology.
Why government services are a good match for AI
Government services are generally not bespoke or customized. There’s very little discretion for the rank and file worker. They’re running through their list of steps as fast as they can. And they do this over and over and over again. An employee reviewing disability claims may iterate on this same workflow once every hour while a 911 operator may do so every 5 minutes. In either event, it’s repetitive.
What’s exciting is that clear SLAs and repetitive workflows lend themselves incredibly well to automation.
Most government workflows also involve forms and documents, something that LLMs are incredibly good at. At the end of the day, any government service is a natural language request sent by the user (constituent) and a natural language response returned by the government.
The market is bigger than you think
And, even though today’s tools are severely lacking, they still point to a huge market for software in government. Tyler Technologies, the market leader in government software, generates nearly $2B of annual revenue per year and is valued at $24B. More ridiculously, Booz Allen Hamilton generated $140m over a 4 year period just for maintaining the recreation.gov website. How much could a company that reduces real administrative burden be worth?
Breaking in: Who to sell to
There’s a few different models for how to streamline how citizens engage with their governments. You can sell directly to the government or you build a brand with consumers. In some cases, there may be a chance to do both.
Selling to the government
Selling to the government is hard. Procurement cycles can be long and buyers can be economically irrational relative to for-profit companies. However, if you can crack government sales, contracts can be large, long, and very sticky. That’s why companies like Tyler Technologies trade for over 11x sales.
Furthermore, a lot of the inefficiency in government services lies with the government itself. You can work with the consumer or a business to streamline their end of the equation but the buck stops with the government.
Selling to consumers
In some cases, a lot of improvement in outcomes can be gained by helping consumers or businesses navigate the hoops that our government makes us jump through. A signal that a market has opportunity to sell to consumers is if there is an existing set of service providers who play this role already.
Examples of this include Social Security & Disability where there is an established industry of law firms who help people navigate these claims and permitting where permit expeditors exist to accelerate construction permitting.
The benefit of this model is that you avoid the government procurement process. Customer acquisition looks more like traditional consumer or B2B sales. The challenge is that companies using this model are still reliant on the government for the ultimate delivery of the service.
Selling to both?
Selling to both consumers and the government is a bit of an unprecedented business model but one that I think the most successful businesses in GovTech will pursue. I suspect the more common wedge will be starting on the consumer side, building a brand, and then using this as leverage with the relevant government institutions.
If your technology can assure a top 1% accuracy disability claim or permit request, perhaps there’s an opportunity to partner with the relevant government institutions to create a “fast lane” of sorts. This would essentially be a separate queue that would have a streamlined review process by the government. This reduces administrative burden and eliminates the risk of government bottlenecking to end-customers (consumers or businesses).
If you’ve seen ways that this has worked in the past (or have ideas), please share them.
Get in touch!
There’s so much opportunity in government services and different problems to tackle. If you’re thinking about how technology and AI can transform the way we interface with our government, please get in touch!
—
Some companies building in GovTech today
Some emerging segments of the market and associated companies. Note that not all of these companies leverage AI today.
Social Security and Disability
Parks & Rec
Constituent CRM
Real estate / permitting
911 / Emergency Services