How we evaluate AI-powered service companies
Where we see the opportunity for venture scale outcomes in the tech enabled services space
For a long time, we’ve been told that software is eating the world. Yet we still pay for services, handsomely and often. In fact, over 63% of Australia's GDP is wrapped up in the service economy1. After all, it’s nice to have something done for you in its entirety, rather than being given yet another tool to do it yourself. Unsurprisingly, there are some really, really large service businesses. Yet, as venture investors, we tend to stick to products, as traditional wisdom dictates that services cannot be scaled. However, many are beginning to ask whether this still holds true in the age of AI, in which complex, non-repetitive tasks are now well within scope for AI models. Can the AI-powered service model fit the mould of venture capital returns, and what might the hallmarks of this new thematic increasingly being referred to as “services as software”?
Traditional SaaS was premised on the idea of selling productivity - amplifying the performance of individual users through software that was largely self-serve. Alternatively, services businesses sell entire, end-to-end outcomes (and the relief of a job done well). The business model underlying SaaS is normally a fixed price per seat; whereas AI-powered services businesses are likely to retain the business model of their non-tech enabled predecessors: one-off fees (or retainers) in exchange for taking ownership of entire tasks or outcomes.
As investors today, we are hunting for problems that can be sold as outcomes instead of software. We are looking for businesses selling them as use cases that are repeatable and scalable; ones that can be done over and over again, to a very high quality and very quickly. From our research and diligence, we have uncovered 8 features shared by AI-powered service businesses that point to the potential of a venture scale outcome. Of course hitting a few of these is challenging. Confidently hitting all of them is rare.
1. Large addressable market
This has to be the starting point for any VC backed business. When we run this over the lens of AI powered service business, we’re looking to triangulate using both ends of the spectrum: existing SaaS spend as well as the market for traditional services businesses. Accounting is a good example of this dynamic in action. Decades ago, Xero and Quikbooks launched, selling accountants tools that made their jobs easier to perform. Both have been undeniable successes - growing into A$20BN and A$200BN companies respectively. Likewise, it’s estimated that the global spend on accounting services exceeds A$800BN2.
Startups like Hnry (automated tax services for self-employed people, delivered by both an AI concierge and accounting experts) and Thriday (automated accounting and tax that shows the health of your business in real-time) are well positioned to pull from both markets. Like traditional services firms, they offer end-to-end accounting services, and like SaaS, they use technology to make it affordable. A noble ambition for both could reasonably be growing to the size of a ‘Big 4’ firm, without the friction of scaling humans and the consequences of the business model attached to that.
Others like Harvey are seeking to straddle both worlds, building a legaltech platform and AI co-pilots to power legacy law firms as they level-up their AI-powered capabilities. Whilst firmly a SaaS tool, co-pilot solutions like Harvey might be seen as ‘double dipping’ from both the existing legal software market, as well as the larger pool of ‘billable hours’ of junior legal professionals like paralegals.
2. Potential for strong gross margins
To be attractive to a VC investor, healthy margins need to be evident and achievable. The archetypal AI powered service business might achieve this through some variation on these dynamics:
Control the flow of information, in order to
Efficiently automate workflows through structured data and repeatable processes, leading to
A lower cost to serve and more scalable backends, and
A superior experience and the right to command a higher price point.
For example, in every engagement, Hnry controls the flow of information via a standardised intake form and on-platform functionality such as expense management and invoice generation. With structured data, preparing documents for submission (like tax returns) is significantly easier. Hnry’s price point can be sizeable, accounting for 1% of users’ annual income. Yet their customers clearly see value for money, whilst a high transaction value leads to more room for margin. Arguably the only downside of the Hnry model is that a trusted advisor relationship is still a critical part of the customer experience - to do this well still requires human touch (for now).
3. High Lifetime Value
This is obtained through both a high transaction frequency as well the right to upsell and cross-sell. PropHero, (a platform that helps you build real estate wealth throughout your life, from finding high performing investment properties to guiding you through each step of the purchasing, financing, and management process) has proven their ability to cross-sell and upsell in their ecosystem. As users move through the journey of purchasing an investment property, PropHero supplements core revenue through their network of mortgage brokers, property surveyors, rental agents (to help secure tenants) and landlord insurers. It’s a tremendous value add to their users to serve up the right people, at the right time, with the stamp of approval from their vetted ecosystem.
“Players who manage to orchestrate their ecosystem and become the "one-stop-shop" for users have the ability to get the best of both worlds: Scale through technology, and high revenue and margins from the services connected. We have a >80% cross-sell rate on all the verticals connected to our platform, as users look for the simplicity of having everything under one roof.” Mickael Roger, Co-Founder and CEO, Prophero
4. Efficient Distribution Model
Rightfully, our excitement regarding the potential of AI-powered services businesses is predominantly driven by the potential for AI to radically transform the product experience. That being said, taking the leap from a traditional services business also requires unlocking new scalable distribution channels.
Whilst ‘cracking go-to-market’ is a truism for all startups, it’s important to highlight here as of particular note for AI powered service businesses. Many scaled service providers (including those taking significant venture funding) have ultimately lost the ‘ground game’ against a field of smaller, leaner ‘mum-and-pop’ providers reliant on largely unscalable, but highly efficient channels - namely word of mouth in a given ‘micro-market’ (e.g. the local suburb). Success requires solving for both scale and efficiency (digital marketing alone is rarely enough) - and a tight understanding of your competitor’s CAC (even where they might be so small they might not even know it themselves).
Practically, a powerful go-to-market motion looks like strong partner referrals, an engaged community (that become influencers and engage in word of mouth), healthy paid marketing economics and an organic SEO engine. Everlab (a preventative health protocol that unlocks access to a premium preventative health service) has these features in play. Their first users are health forward tech enthusiasts and are vocal advocates for the brand. Everlab has wrapped its arms around this longevity community and invited them to be part of championing a new approach to health. They activate this community via viral loops in their product - users proudly share and discuss their Everlab diagnostic results on forums and social media. Ultimately, this sticky, badge-worthy protocol is the backbone for their efficient distribution model via referrals and word of mouth.
In a similar vein, PropHero also benefits from scalable referral dynamics. With consistent cadence, other marketplace providers are compelled to send customers PropHero’s way. Take mortgage brokers, who want to move approved applicants through the funnel by connecting them with tools that accelerate a purchase - like PropHero. Matilda Migration (an AI-powered digital migration agent), similarly reap the benefits of consistent referrals. Demand for visas is high and often concentrated in particular segments - like education and spousal visas; partnering with existing aggregation points such as Enrola (an education comparison tool for vocational courses) presents an elegant referral channel for Matilda.
5. Market Dynamics
The best kind of market dynamics are fragmented markets with a long tail of smaller players, and no clear market leader. In services businesses especially, fragmentation can be correlated with low NPS, as standardisation across the industry is poor and players are less likely to differentiate on superior customer service versus alternative moats such as local distribution moats (e.g. being the only migration agent in the neighbourhood). Ultimately, this creates an attractive welcome mat for new arrivals. Conversely, a highly concentrated competitor base would be reluctant to allow a new intermediary into their market, and would be ready to fight the threat to their revenue by forcing prices down and squeezing margins. We need look no further than the travel industry to see this play out. The large airlines have all but obliterated the economics of online ticketing marketplaces, leading online travel agents to focus on hotels where fragmentation, and therefore the economics, are higher.
Matilda are entering an optimal market dynamic, with an approachable service that guides migrants through the visa process simply and efficiently. In their industry, there is no clear market leader, with fragmented competition dotted throughout the landscape. This paves the way for a smooth and largely uncontentious entrance.
6. Customer Dynamics
Successful AI powered service businesses are poised for growth when NPS is low, yet the emotional stakes surrounding the transaction are high. In the case of Matilda, we see this dynamic firmly at play. It’s hard to put into words how stressful, overwhelming and intimidating the process of migration can be; starting with the paperwork. Lives, and the lives of families, hang in the balance waiting for critical documentation to process. Yet a staggering 33% of migration agents receive complaints3. With this as a backdrop, Matilda entered the market. Their AI powered service automates much of the manual process surrounding visa documentation; reserving the human touch for critical moments. Due to the emotionality of the transaction, there is a potential that there will remain a preference to interact with a human for more than a few fleeting engagements. However, this can be overcome with proven parts of the DTC marketing playbook; particularly reviews, user generated content and product guarantees.
“Often the complexity sits at the start of the visa process, trying to understand which visa is appropriate for their circumstances. This is also, ordinarily, where people want support and guidance (and therefore human intervention). Once their circumstances have been assessed and a plan decided, the process is relatively straightforward - automation and AI supports the client to build their application.” Niamh Mooney, Co-Founder of Matilda
7. Potential for Cornered Resource
When a business has exclusive access to a product or resource, via IP, their team, a distribution strategy or marketing engine, it is a positive signal that a business could be on the venture track. In the case of AI powered service businesses, this means ensuring any cornered resource can equally scale alongside the business
Take PropHero, their streamlined investment process creates immense value for not only buyers (their customers), but sellers as well - namely seller’s agents who are incentivised to minimise time on market provided they can find a fair price for the property. Hence, PropHero have quickly become the first place seller’s agents approach with new stock, not yet available to the rest of the market, further cementing a network effect and ensuring PropHero are able to adequately scale supply alongside demand.
8. Competitive Dynamics
Here, we’re looking at whether competition is well serviced by existing technology or not. Ideally, there should be an existing predisposition to technology, as well as some kind of moat to help avoid fast followers. In other words, the new entrant would choose a category where there is already familiarity with a tech-enabled solution and then design an AI-powered system that gives them a structural advantage over their traditional competitors. Predisposition to technology is critical here. In some categories, a human interface still matters. For example, while parts of the legal process can be accelerated with AI, there is still comfort in knowing that a person is on the hook for advice given. In the instance of PropHero, comfortability with browsing real estate online is already high thanks to the likes of REA and Domain, albeit low when it comes to actually transacting (creating a gap for the team to neatly fill).
In summary, It’s no secret that AI has reduced barriers for entry. And while that’s good for startups looking to explore the idea maze, it’s also good for incumbents - arguably more so as they already have the advantages that scale brings; namely distribution, an existing customer base to cross sell to, and proprietary data built from years of operations. Ultimately, the fundamentals of building a venture scale business still matter and are unchanged - perhaps evident given the ‘universal’ nature of the above 8 factors. It’s not enough to just be an AI business, without finding a true competitive moat. If you have thoughts to add into the mix, we’d love to hear from you.