Why AgentGPT Could Be A Big Deal
Based on OpenAI’s GPT-4, AgentGPT, launched in mid-April, can be used to create personal digital agents that attempt to solve user-assigned goals. Such goals can be anything from creating a marketing plan to managing an investment portfolio or building a website. The digital agent seeks to solve these goals by breaking them down into subtasks. It then applies feedback loops to optimize the solution. To highlight why this could be such a big deal, let’s look at the simple example of booking a family trip online, today and in the age of agents.
- Booking a family trip yourself. Most likely you compare prices on multiple platforms, read reviews on others, and try to figure out hidden costs and insurance fees of car bookings and whether they can provide a baby seat. Which airline will accept your loyalty points and how much do they charge for a three-year-old? Do any of the seaside hotels offer a crib free of charge? Which hotels offer gluten-free breakfast and are close to golf links?
- Booking with a digital agent. With natural language prompts, you tell your digital agent to book a trip for two adults and a baby to your destination of choice. You provide criteria for hotels, cars, and activities. The agent creates subtasks for each, checks maps, and analyzes pictures, reviews, FAQs, and T&Cs in more than 120 languages. It presents options back to you, asking for further input if needed. The agent optimizes until you’re happy with the result.
This basic scenario highlights how much friction still exists in the online world. Digital agents have the potential to remove these frictions and fundamentally change how clients access the internet and interact with digital businesses.
Rewriting The Rules Of Digital Strategy
Marketplace business models have been the winning strategies in digital business to date. Retail, travel, ride hailing, and personal entertainment are dominated by marketplaces such as Amazon, Expedia, Uber, and YouTube. How would these strategies fare in the age of agents?
- There go my network effects. Network effects were long considered the holy grail of digital strategy — virtuous cycles of more clients attracting more suppliers and vice versa, leading to quasi-monopolies. While clients appreciate the convenience provided by supply aggregation and ecosystem orchestration, a digital agent can aggregate significantly more supply than any “man-made” platform business. It can do so across language barriers and filter for the most relevant attributes in seconds.
- Now you’re just somebody that I used to know. Digital marketplaces enjoy frequent client interaction due to their aggregated supply advantage over individual retailers. This enables them to curate their offering based on the rich data they harvest from these interactions. If the primary client relationship shifts to a digital agent, marketplaces lose their curation advantage. While agents would have to supply some amount of client data to the marketplace, a digital agent won’t be susceptible to clicking on the next best offer.
- The end of price discrimination. Many marketplaces have become de facto price search engines in their sector, making competitor prices irrelevant. Exploiting this, digital marketplaces oftentimes engage in price discrimination, asking different prices for an identical good from different customers. If digital agents compare prices from an endless number of suppliers, large gateway suppliers won’t be able to fool their customers into overpaying as easily anymore.
Looking at these scenarios, digital agents have the potential to create a world in which supply aggregation is meaningless, curation goes into thin air, and everybody is always multihoming — a cascade of nightmares for existing digital marketplaces.
Don’t Count Out The Incumbents
If agents make online purchasing decisions on behalf of humans, existing marketplaces could see their business models under severe pressure. They have several options to react to such a trend.
- We reserve the right to refuse service to any bot. A defensive reaction of existing digital businesses could entail doubling down on existing bot barriers and moving content behind paywalls or membership walls. However, this approach will become increasingly difficult with more and more intelligent agents equipped with client login data.
- The rise of agent experience. Taking the offense, marketplaces could embrace AI agents as a key customer group. Content providers have been optimizing their content to be readable for Google’s search algorithm for quite some time. Similarly, digital marketplaces could seek to drive agent experience to make their offering convenient from an agent perspective.
- Would you like a movie with this? “When we win a Golden Globe, it helps us sell more shoes,” Jeff Bezos famously said about Amazon’s movie productions. Marketplaces often offer supplementary benefits to keep clients tied to their platform. Identifying such benefits in the age of agents by offering ancillary, undeniable value to (human) clients could therefore be an antidote to losing the primary client relationship to people’s digital agents.
- Try to climb this wall of fixed costs. Barriers to entry are extremely low in digital businesses. Realizing this, many digital giants have created fixed cost moats that provide value to clients and act as a deterrent to potential market entrants. Company-owned brands, logistics networks, exclusive content, and proprietary consumer devices could become even more important for incumbents in a potential age of agents.
- The race for agent supremacy. Tools like AgentGPT can supercharge existing, corporate-designed virtual assistants. Equipped with the funds and computing power to create easy-to-use and well-trained digital agents, corporations will try to lure consumers to their closed-code digital agents. Consumers might opt to sacrifice autonomy for convenience, turning digital agents into a client interaction layer that continues to optimize business interest.
This post was written by Principal Analyst Manuel Geitz and it originally appeared here.