AI is coming for your (services) job
☠️

AI is coming for your (services) job

At 1984 Ventures we have been investing in antiquated industries since 2018. For a few years these sectors were overlooked by investors, but then caught a huge tailwind with Covid. What’s now clear is that the tailwind provided by LLMs is an order of magnitude more disruptive than that of Covid. LLMs are transforming industries that were historically uninvestable, including services industrieslike recruiting, bookkeeping, market research and government contracting. More provocatively, the less historically interesting the industry from a VC perspective, the better a place to build an AI startup today.

Why AI will eat the services sector

There are a few reasons why such industries are now fertile grounds for innovation:

  1. No software incumbents: Services industries have historically been bad places to invest for a variety of reasons: buyers lacked software budgets, high fragmentation, software was more of a nice-to-have than need-to-have etc. As a result, many of these industries lack a dominant software incumbent. Since a key tenant of AI investing is avoiding head-to-head competition with an incumbent’s distribution advantage, this provides critical air cover for startups
  2. Labor intensive + workflows ripe for automation: Almost by definition, services industries are labor intensive. Tasks like answering phone calls, scheduling appointments, doing research, updating policy documents, drafting memos, responding to emails and doing outbound are commonplace.

💡
Recent technological advancements in LLMs, computer vision and text-to-speech/speech-to-text have reached a tipping point: they are able to substantially outperform humans when undertaking many of these tasks
  1. Prime target for selling work, not just software: The combination of labor intensity and the high propensity for workflow automation makes these industries a prime candidate for selling work, not software. Said another way, the AI software’s output can be easily compared to the units of labor that previously undertook the task. This allows the software to be priced substantially higher than regular SaaS.
💡
Given this paradigm shift in pricing and the radical efficiency gains AI can provide to employees and businesses, there is potential to completely rethink (and radically expand) markets. Sectors that may have previously been considered niche can suddenly have venture-scale potential (see examples below)
  1. Ability to massively change the industry’s margin profile: Many of these industries have thin margins given the labor centricity of their businesses (e.g. the recruiting industry has 5% net income margins, auto dealers have 2.5% net income margins). The value prop of using AI to drastically increase the leverage of the workforce (or radically reduce the workforce) is attractive to business owners, who can easily visualize its impact on their bottom line
  2. Willingness to adopt the tech: Surprisingly, even though these industries are not historically huge buyers of software, they have demonstrated a willingness to adopt AI. This is because the value propositions are so obviously better (10x…sometimes 100x) than the status quo. We have seen this play out with various portfolio companies such as Revv (with auto body shops embracing AI software), Seso (with farms adopting payroll and workforce management software) and BuildOps (with commercial contractors using business management tooling)
Note that some services jobs/tasks are easier to automate than others by virtue of their inherent complexity. Additionally, the higher the frequency of the task, the clearer the value for automation. And finally, in some cases the service provider’s product is not simply the completion of a task but also the stamp of legitimacy and approval (for example, an auditor). In those cases, an AI alternative doesn’t just have to have comparable output but also needs to match the trust factor, which can hamper adoption.

The above may sound like a nice theory, but I think some examples are what really brings it to life.

Some examples

  • InspectMind (AI that drafts inspection reports)— when buildings are built you typically commission a construction inspection. As part of that process, inspectors go on site, take notes + photos and then go off to write a report (which can be several hundred pages and takes many days to produce). With InspectMind, an inspector takes a video and provides a voiceover while touring the site. The AI uses these inputs to automatically draft a 200 page report. This saves the inspector days and most importantly, allows them to massively increase the number of inspections they undertake. Construction inspection firms (like that of the founder’s family) are typically 1-2 person SMBs with low profit margins; asa result, they haven’t historically been great buyers of software. But since InspectMind enables these firms to undertake more jobs and make more money, they can command sizable ACVs (priced according to revenue uplift). If InspectMind can find a way to hack distribution, there is an opportunity to build the next Procore ($11B market cap) focused on SMB firms
  • Apriora (AI recruiter) (*) — recruiting has historically been a poor category for software as in-house recruiters typically don’t have large budgets (and software tools have’t radically changed the job of a recruiter). But by choosing to sell work, not software, Apriora has the potential to reinvent the industry. Instead of building software for recruiters, Apriora IS the recruiter. They screen CVs, conduct voice interviews and then synthesize the calls (displaying humanlike judgment). As the website says: create your interviewer on a Friday afternoon and have interviewed everyone by Monday morning. The output of their software can be directly compared to the amount (and cost) of labor it previously took to conduct the same number of interviews. As a result, the ACVs they can command are substantial
image

A screenshot of Apriora’s product, including a snapshot of the interview ‘score’, which summarizes and synthesizes feedback on the candidate.

  • Artos (AI for scientific draft submissions) — Life sciences companies spend many months writing documents as part of regulatory submissions, which are often the final hurdle before a drug can be brought to market. Using Artos’ document writing and synthesis tool, high quality drafts can be written in minutes rather than months, shaving off drastic amount of time and hundreds of thousands of dollars in the process

Concluding thoughts

What the examples above demonstrate to me is that category creation and reinvention is occurring across industries that investors would have historically avoided or overlooked. Industries like recruiting, accounting and market research were considered far too labor-heavy to be interesting. However it is precisely because they are labor heavy that there is a sizable opportunity for automation. Automation should substantially increase efficiency and lower the cost, which has the potential to meaningfully expand the size and scope of these markets.

Despite my clear optimism, I’d be remiss to end this post without the following disclaimer. Finding a good starting point in an overlooked industry doesn’t guarantee you’ll create a great business. If anything, both 10x ideas and 10x the competition now abound. As a result, it’s important to consider from the outset how you’ll hack GTM and build defensibility into your product (through things like integrations, partnerships, product expansion, proprietary technological developments outside of AI etc). But if you can figure those things out, you may be one of the entrepreneurs whose name is etched in the annals of history, as this may be the beginning of a once-in-a-generation opportunity to transform massive swathes of the economy.

If the above post resonates with you and you are building an AI startup in a legacy industry, I’d love to hear from you (I’m at samit@1984.vc).

* 1984.vc portfolio company