Self-service and the future of software

Developers write code. LLMs are transforming how code is written. We can all agree on that much, right? Even if we don't know how it will ultimately look: whether LLMs will serve as copilots inside IDEs, or if they will "compile" plain-English requests into source code, and 'software engineering' will consist mostly of reviewing that LLM-generated code.

...But code is only one of the two outputs engineers create. The other — often larger, by volume — is communications about code. Commit messages, Jira tickets, PR reviews, Figma comments, Notion documents, Slack conversations, emails, standups, meetings, reports, documentation, test plans, etc. These are the oil in a project's engine, the reason teams can work efficiently, in alignment, and in accord with their managers' directives. If they aren't there, or aren't good ... the project sputters, even backfires, and much of the time and money poured into it turns to wasted entropy.

Our fundamental thesis at Dispatch AI is that LLMs will revolutionize these communications about code no less than they will transform code generation itself.

As such we've automated maybe their most important task: reporting on software teams' progress, status, and risks to managers and execs. It's absurdly easy to set up — just take 15 minutes to connect your tools (GitHub, Jira, Notion, Figma, Slack, Sentry, etc. etc.) and we start sending you weekly detailed & insightful reports, no chatting required. (You can ask followup questions, of course.)

In fact it's so absurdly easy that this week we're softly launching a 100% self-service version. Log in at https://thedispatch.ai/config/account; sign up for our 21-day free trial, no credit card required; connect your tools, and start receiving reports. If you have the sneaking suspicion that your software project(s) could be running smoother, give us a try. I'm confident you'll be impressed by our reports' sharpness and accuracy.

That said, these are very early days. This is the MS-DOS era of AI. Dispatch would have been impossible to build only 18 months ago! Current frontier models are amazing, but I can't wait for the next generation. Today's LLMs understand software best, so code is what they're best at; and just as over the last 20 years, automated testing became table stakes — kids today don't realize that back in the oughts, engineers ~never wrote unit tests! — LLM-powered analysis and reporting will become essential for software teams everywhere. That's a given with today's models alone.

The next generation, though? They will be just as good at other industries, from sales to media to construction. Ultimately, and sooner than you might think, automated analysis will be taken for granted at every organization worldwide, providing a second set of eyes, and a safety net, for everyone. That, I submit, is a very big deal.