By Güimar Vaca Sittic and Fabrice Grinda

Online marketplaces are fascinating. Matching demand and supply seems mundane, but it takes a lot of work to actually make it happen.

As we pointed out before in The Evolution of Marketplaces, marketplaces have evolved dramatically from horizontal listing based models like Craigslist to vertical transactional end-to-end marketplaces like Uber today.

Every few years there are technological breakthroughs that allow entrepreneurs to adjust the dynamics of online marketplaces to make them more efficient. Native mobile apps, seamless payments and GPS tracking are just a few of the examples that have improved marketplaces substantially during the last two decades. Today, we are at the beginning of the next evolution: AI-powered marketplaces.

Every marketplace has a chicken and egg problem. Ultimately ever more buyers bring ever more sellers and ever sellers bring ever more buyers, but it’s hard to get the dynamic started. We recommend building supply first as they are financially motivated to be on the platform. The nature of the marketplace dictates the ratio between supply and demand. This ratio varies based on the relationship between buyer and seller (i) one to one (e.g.; Care.com), (ii) one to many (e.g.; Udemy) or (iii) many to one (e.g.; Uber). This ratio often dictates unit economics and how attractive the business is. AI powered marketplaces can change this dynamic.

When talking about AI, it is important to note the difference between AGI (Artificial General Intelligence) and ANI (Artificial Narrow Intelligence). AGI refers to a computer which is as smart as a human being across every single category or task. AGI would include not only problem solving but also abstract thinking, ability to reason, etc. AGI is the AI of Ex Machina, Her and Jarvis. It is still years away. ANI is already present across many different specific tasks. ANI can be as narrow as your spam filter. Gmail learns from your behavior and those of the community what emails are not desirable and sends them directly to spam.

Small AI-powered tasks such as your spam filter might seem insignificant but the aggregation of many ANI applications can be transformative. Many startups like X.ai and Clara Labs have been trying to improve a basic but time consuming task like scheduling. They use a combination of Natural Language Processing (NLP) and machine learning to reduce the number of touch points needed by a human being. Until recently, busy individuals needed to have an assistant to manage their schedules. The assistant checks availability, shares it with other meeting participants, finds convenient times, sets the meeting and sends a reminder. Some parts of the chain need good judgement, but others like sending reminders can be automated. By automating many tasks, the assistant can manage the schedule of several busy individuals rather than one radically decreasing the cost of the service. This can be applicable in many types of marketplaces creating the opportunity for a triple win where consumers pay less while suppliers and the marketplace make more.

There are two big groups of marketplaces: product-based ones and service-based ones. Product-based marketplaces can use AI to improve their internal processes and decrease fixed costs. For instance, AI powered back office tools can improve quality control and customer care. AI can also be used to dramatically improve user experience. AI could identify pictures and suggest descriptions and prices radically simplifying the selling experience.

AI can be even more potent for knowledge-based service marketplaces. There are many marketplaces where the service provided by a human being that has deep knowledge and expertise in one particular topic. UpCounsel is such an example. UpCounsel is a fantastic marketplace that connect high quality lawyers with customers who need their services. If you want to file for a patent and don’t know where to start, you should go to UpCounsel and consult with a lawyer. The lawyer you just hired will certainly be very good but he also needs to pay for his bills and getting a J.D. can cost up to $200,000 in top tier law schools. Thus, hourly rates for great lawyers vary from $150 to $500. Filling your patent might take 10 hours at $200 per hour costing a total of $2000. In the (not so distant) future, UpCounsel can build ANI tools for their lawyers that integrate in their work flow to decrease the amount of work needed from the human being. An ANI tool might decrease the time needed by a lawyer to check if your idea/product qualifies for patent protection. If such tools yield a 3x productivity in time for a lawyer then the marketplace can charge less per hour, yet earn more for lawyers, while having a bigger take rate.

Example:

Imagine you have to file for a patent and need 10 hours from your lawyer to complete the task. The hourly cost for the end consumer is $200.

Scenario A (without ANI):
Cost to consumer: $2,000
Lawyer revenue: $1,600
UpCounsel Revenue: $400 (20% take rate)

Now imagine that same lawyer was a 3x productivity boost and can serve 3 clients instead than 1 during those 10 hours of work.

Scenario B (with ANI with 3x yield):
Cost to each consumer: $1,500 (25% discount)
Lawyer revenue: $3,000 (90% increase)
UpCounsel Revenue: $1,500 (350% increase, 33% take rate)

A lawyer working full time for UpCounsel might serve 20 clients per month. That’s a 20 to 1 ratio between supply and demand. Now imagine what it could be with 60:1 or even 600:1.

There are many categories that can be disrupted by ANI such as tutoring, health, law, mental health, nutrition, programming, transcription, translation and countless others. If you slide and dissect each one of these services into micro tasks you will find out that many of them can be automated by building ANI products. The jury is out on whether the current marketplaces will be smart enough to adapt to these changes or if there will be newer bolder entrants that will leverage ANI to outcompete the current incumbents.

Either way, the future belongs to AI powered marketplaces!