Why Selling Outputs Falls Short Without a Human Touch
After being considered a “no fly zone” for the past few decades, services businesses are suddenly in vogue in Silicon Valley. By services, we don’t mean the wave of consumer-oriented utility businesses in food, transportation, or home maintenance. Instead, we’re talking about knowledge-based work such as legal, financial, and healthcare services. At Forerunner, we’ve seen more businesses in the tax space in the past year than we have in the prior decade.
The first wave of software-enabled services businesses emerged over a decade ago, driven by new GPS capabilities and the widespread adoption of portable, internet-connected devices across all demographics. Now, a new wave is being catalyzed by AI, which automates knowledge tasks, in contrast to the physical tasks automated by the previous generation.
Each wave is first addressed by software companies filling in the gaps of adoption. In the early 2000s, pioneering two-factor authentication companies helped banks authenticate logins by sending codes to users’ mobile phones. Now, receiving six-digit codes for logins is so commonplace that iMessage auto-populates them, but companies that introduced this technology no longer exist.
In fact, not only was that service commoditized, but Twilio, the company that popularized it as part of its broader suite of telephony solutions, is now facing a similar margin squeeze on its own core business and falling victim to this same cycle. Meanwhile, the consumers of these solutions — pioneering services businesses like DoorDash, which reinvented delivery, and Uber, which reinvented transportation — are thriving. As technologies mature, you typically need to either be the underlying infrastructure or deliver the complete service. Middlemen offering solutions too often get squeezed.
When it comes to AI, the solutions selling into services firms have the potential to handle a much higher percentage of tasks than traditional workflow or rules-based systems, but many of the industries still require a human touch. Some services, such as law, healthcare, and architecture, must be delivered by a credentialed individual to be valid. Others, like investment banking and recruiting, demand a high level of trust and confidence due to the nature of the transaction.
In our view, many future winners will look more like historical offline stalwarts of the services world than the sexier SaaS-powered solutions that have dominated the past decade. That’s because — as technologies mature and AI-driven knowledge work moves up the stack to more nuanced tasks — there will be far less proprietary value in selling AI-generated outputs into firms, and thus far greater incentive for new firms to centralize AI productivity in-house alongside the irreplaceable, uniquely human aspects of the business.
Put another way, don’t sell research tools to Goldman Sachs; be the next Goldman Sachs. Don’t sell CAD tools to Gensler; be the next Gensler. Don’t sell analytics to WPP; be the next WPP.
Now, with the emergence of LLMs, Forerunner is doubling down on opportunities that pair the IQ of AI with the EQ of actual people.
When “Selling the Work” Stops Working
Today, generating useful outputs from LLMs requires a rare combination of prompt engineering, model routing, and fine-tuning skills. As a result, highly profitable software companies are being born as contract manufacturers for knowledge work, selling AI-generated components to service providers who used to produce them in-house with human labor.
EvenUp has become a popular example of this model. Instead of helping injury lawyers manage their workflows, EvenUp leverages AI to transform medical records into demand packages – helping lawyers win more cases, settle faster, and save both time and money.
There’s a lot to like about this approach. By selling the work, a new concept that’s driven fresh thinking about AI applications, you can price relative to replacing labor rather than incrementally improving productivity. You also don’t have to deal with the liability, licensure, and overhead that comes with being a personal injury law firm.
But “selling the work” comes with two major downsides:
- As AI advances, the core value proposition loses value as high-quality outputs become easier to produce without supplemental engineering.
- The arbitrage opportunity that currently exists will collapse when prices begin to reflect the growing set of other AI-generated alternatives instead of human labor.
For hints at where we’re heading, look no further than the professional exam performance and 60x cost differences between GPT-4 and GPT-3.5 Turbo. The voice capabilities of GPT-4o close the gap even more. Soon, it’s inevitable that model outputs will become commonplace, meaning service providers themselves – not middlemen – will hold the leverage.
After all, it’s service providers who have the credibility and licensure to sign off on finished work – and the expert judgment and relationships to cut through the noise that AI will inevitably amplify.
The Problem with Legacy Service Providers
If AI-generated outputs become commoditized and services firms themselves hold the power, wouldn’t firms who have decades of experience dominate the AI opportunity? The reality is that legacy providers are very unlikely to win, even if they’re in pole position. That’s for four reasons:
- Low technical aptitude. They’ve historically been slow to embrace and maximize cutting-edge technology.
- Misaligned incentives. Their revenue models are often tied to hours of human labor rather than outputs or outcomes, which disincentivizes productivity gains.
- Outdated org charts. These organizations were built around a different labor model. It is much more difficult to cut staff due to innovation than to hire differently.
- Slow and steady culture. Their company cultures are often incompatible with the pace and goals of an innovative, fast-growing company.
Therein lies the potential for a new generation of AI-native services businesses. New services firms can prioritize AI implementation at every stage, organize expert employees to spend time on work that LLMs can’t replace, align business model incentives in more productive ways, and establish cultures that are far more risk-tolerant and aspirational than their predecessors.
In the process, these companies can also prove critics of the AI + human model wrong by ensuring services are high-margin, sticky, and scalable – all while traditional seat-based SaaS businesses are having their margins rethought in the face of continuous R&D, intense competition, and new inference costs.
Why the “People Business” Will Become Even More Valuable with AI
Much attention has been paid to the roles AI will replace, but there's been less discussion about where it will make people even more valuable. And it clearly will make some people more valuable – particularly on the revenue-generating side of the business where experts can leverage hard-earned experience and relationships to cut through noise.
In knowledge-driven services businesses, the most senior people in the organization spend as much time generating business as they do running it. Take McKinsey as an extreme example. Making partner rarely happens because of phenomenal analysis; it’s more likely one gets there by signing and retaining strategic clients.
Now, think about an AI-enabled McKinsey. Not only could the associates research projects more efficiently, creating beautiful presentations in mere minutes, but senior partners could sell more work through assisted outreach and dynamically-generated proposals based on specific client needs.
We’ve seen this recently at our portfolio company, MakerSights. They historically sold tools to merchant teams at apparel brands to help them better understand consumer demand, manufacture the right products, reduce waste, and improve margins. However, as their customers cut headcount and tightened budgets, the Company faced headwinds that impacted the lion’s share of SaaS tools selling into leading apparel businesses.
Fast forward to Q1 this year and Makersights completely revamped their approach, functioning more like an AI-powered services firm than a SaaS solution. The company now uses their own software combined with consumer panels to analyze a brand’s new collection providing c-level actionable insights, so time previously spent training customers on their product is now time spent delivering direct value. With these tools underpinned by AI, the Founders are freed up to be back in the room with decision-makers and drive customer growth.
While the company might look more like a consulting firm from the outside, they have margins like a software business and grew more in Q1 this year than they did all of last year. Even if the bulk of the work behind the scenes ends up being carried out by AI, Makersights is now much more of a “people business” — exercising judgment, building trust, and adding EQ to effectively communicate outputs.
Zooming out, this is an example of how AI will impact companies’ human labor investments in seemingly contradictory ways. More work completed through automation will require more expertise to validate decisions, build trust around that work, and oversee the expanded capacity. This is not a paradox; it’s natural for companies to invest more, not less, in revenue-generating assets as they become more productive.
Where to Build an AI-Native Services Firm
Not all services categories are the right fit for the hybrid human + AI model. First, any eligible category should pass a simple Goldilocks test: the bulk of the core workflow must be highly automatable, but higher stakes elements like the sales, support, and service quality must benefit from the judgment and relationship skills of people.
Consider the difference between wealth management and financial analysis. Both fields involve calculations that can be fully automated, but wealth management also incorporates a complex emotional dimension. Decisions like how much to save for a child’s college education or whether to help a friend in financial distress are not just about numbers; they’re the kind of nuanced situations best discussed with an empathetic expert.
At the opposite end of the spectrum are services that defy easy automation. Some face physical limitations since AI cannot fix a roof or tend to wounds. Others resist automation because human connection is the core service offered, like childcare. While AI can support learning, it cannot replace the development benefits students get with teachers and classmates at school.
After passing this initial test, the best AI-native services categories also meet three additional criteria:
- Regulatory moat. They benefit from licensure requirements and other regulatory hurdles, which act as a natural barrier to entry and provide a stamp of trust and credibility.
- High trust. They require a high degree of human trust, where a specific person can provide personalized support and be held accountable if things go wrong.
- Pricing unlock. Their cost structure flexibility can overturn the legacy business model, leveraging newfound efficiencies to align incentives more effectively.
The best categories don't just have workflows that can be mostly (but not entirely) replaced by AI. They also offer natural barriers to entry, an economic premium for owning the customer relationship, and a pricing structure that's predicated on human labor and can be flipped on its head. Many legal, financial, and medical services stand out as examples that have all three.
These traits make it possible for AI-native firms to directly target large markets that have been resistant to change, moving beyond previous attempts with workflow management software suppliers or providers of secondary services that didn’t ultimately transform those core markets.
Again, consider wealth management, a highly regulated sector often viewed as outdated and distinct from tech-forward options like online stock trading. Although many robo-advisors have tried to challenge traditional firms, they've struggled to impact the larger market since consumers prefer having a personal relationship with their money manager – even if it’s more expensive. As a result, there’s still a massive gap in scale between offline and online options; UBS manages $5.7T in client assets while Wealthfront only oversees $50B, most of which was acquired through in-person meetings versus online sign up.
Now, there’s finally a catalyst for change in wealth management. New AI-native firms have the opportunity to reimagine the core category by offering even higher quality services than offline incumbents for less, flipping the revenue model from a percentage of AUM to a more modest fixed membership fee while opening the door for a wider set of folks to have their own wealth manager.
Our portfolio company, Atticus, has a similar opportunity in consumer legal services as the costs to serve clients come down through increased AI automation. As a business, Atticus acts as a social security law firm representing individuals with disabilities that make it difficult for them to continue working. Since firms in this category operate on a contingency basis, many cases are not picked up because the probability of winning isn’t high enough to justify the attorney cost to take the case to court. With a dramatically lower cost to qualify and serve, Atticus has the opportunity to take on more cases, growing the business faster while helping more people along the way.
What Happens Next?
Rather than simply transitioning SaaS to selling outcomes into services firms or trusting services incumbents to do the AI opportunity justice, we believe a new class of companies will deliver on this potential by building full-stack, AI-native services firms that emphasize a human touch. These companies will transform larger markets by replacing incumbents, not just selling to them. Importantly, we believe this is what the customer wants: intuitive, efficient experiences with a trusted point of contact.
Put another way, don’t sell research tools to Goldman Sachs; be the next Goldman Sachs. Don’t sell CAD tools to Gensler; be the next Gensler. Don’t sell analytics to WPP; be the next WPP.
These markets are still dominated by the same organizations as decades ago, but that won’t remain true forever. We look forward to partnering with bold teams ready to change the status quo, embracing each wave of new technology while ensuring the face of these industries remains as human as ever.