The Rise of Intelligent Agents
Intelligent agents are emerging as serious contenders to pursue the holy grail of simplifying and enhancing users’ lives by being a high value-add concierge. Agents are user-centric, transform data to insights and actions, and focus on delivering specific outcomes. They come embodied as apps, bots, add-ons, APIs, or even special purpose devices, but their essence is the disembodied intelligence powered by big data that helps users holistically navigate both their virtual and physical worlds.
We are seeing the rapid emergence of four types of intelligent agents.
Platform agents: Owned by the Mobile OS, a messaging platform, search engine, or an e-commerce platform.
Supplier agents: Owned by a supplier — a company who sells products, services or content — like Sephora, Whole Foods or The Weather channel. They can be deployed on multiple platforms as apps, bots, web sites, or an API.
Independent agents: Owned by third parties, who are not makers, but are connected to a network of suppliers, web sites, services or apps. (Like Zeel, Thumbtack or Uber.)
Autonomous agents: Created by a third party, but deployed and managed by the end-user — meaning the consumer or the customer. In their own or managed environments. (Like ad blocking add-ons.)
Some agents transcend many categories. And act like super agents.
Like Google Now - which is an independent agent that orchestrates and pulls content from multiple services and lives on multiple platforms; it is also a supplier agent that offers maps, search, youtube and many productivity tools; it is also a platform agent on Android (Google Assistant) with access to and knowledge of everything related to the user.
Or Alexa, which is a supplier agent selling Amazon products, while also being a platform agent (has its own OS and can live across multiple devices), and an independent agent (connects and orchestrates other services).
Arguably, a lot of this is already happening today.
But you ain’t seen nothing yet.
The number of these agents and their adoption is set to explode as they become more cognitive and intelligent.
The business of agents will be fueled by closed loop data science operating on massive troves of user data — aggregated, anonymized, temporalized and contextualized — to deliver specific insights and enable effective actions. Agents will have simpler, sometimes non-existent front-ends, and use data and intelligence to drive better outcomes for users.
The business of agents also has the potential to dramatically reshape search, advertising, lead generation and customer engagement. Especially as the dollars continue to shift to mobile, creating natural inflection points.
How Will Agents Make Our Lives Better?
Platform Agents
Focused on productivity today, platform agents are likely to expand into commerce, content, entertainment, travel and much more. They will support both push and pull modes.
Push — Curate lists of apps, products, services and content relevant to the user, and surface them at the right times using contextual intelligence.
Pull — Provide results to user requests for specific jobs to be done. And actualize by having the ability to fulfill or provision them.
Platform agents will seek to curate and mediate across multiple installed apps, yet-to-be installed apps, and APIs. They will execute services in the background, install apps if necessary, and orchestrate the experience.
Platform agents will be omnipresent on any logged-in device running on the platform and will operate in multiple interaction contexts — voice, search, messaging, cards and others.
Platform agents work for and on behalf of the user, but in the long-term, they have the potential to corral a significant share of mobile advertising budgets, both around direct response and brand advertising. Especially as more people use them as the gateway to the internet, and to find, know and do stuff relevant to their lives.
Supplier Agents
These belong to a specific company. They can be deployed on public app stores, bot stores and web sites, or even as headless APIs. They can also be deployed on restricted private stores and used by employees, resellers, dealers and customers.
The goal of supplier agents is to help their stakeholders engage with the company faster, simpler and easier. From their perspective, not the company’s. And for the company to realize key insights about specific products, customers and business processes.
Consider the case of a heavy equipment OEM who builds an advisory agent to help its dealers sell more effectively to farmers using tablets. It uses past transactional data of the dealer and anonymized transactional data of other dealer transactions with similar customers to deliver targeted recommendations and pricing.
Or a beauty products retailer, who deploys a smart agent on a messenger platform. The agent can analyze a selfie image to suggest the most relevant products to a female prospect. The agent extracts metadata from the image — facial characteristics, skin type etc. — and uses machine learning from a massive dataset to hone in on the most appropriate products.
Or a grocery retailer agent who — with your consent — gets access to the contents of your smart refrigerator by partnering with the manufacturer, and to re-order and deliver your grocery, just before they run out.
We will see rapid emergence of platforms and toolkits that allow suppliers to build and deploy agents. Not just the front-end, but also the data ingestion with prepackaged integrations, predictive and prescriptive analytics, and the ability to execute specific actions.
Independent Agents
Independent agents sit atop platforms that connect customers to a network of suppliers.
Independent agents typically request access to some of your sensors or data from cloud services, and in turn, help users become more productive, find suppliers and get their jobs done more effectively.
Independent agents will likely be specific to the type of job that needs to be done. They will be optimized for use cases structured around specific outcomes. They will ingest data across multiple suppliers and customers, apply machine learning to realize key insights and predict specific actions.
There may be specific agents to address classes of consumer jobs-to-be-done in transportation, food, dating, education, healthcare etc. Or to provide insights to users in specific professional categories like sales, HR, service and marketing.
They could be targeted to specific communities of users — diabetes patients, inside sales personnel, fitness freaks etc. Or address high-value, intelligence-driven industry use cases like water leakage in utilities, proactive recharge management in prepaid mobile, or real-time cross-sell in financial services.
Consider the case of insidesales.com which makes the function of inside sales dramatically more effective by providing actionable insights to its users on who to call, when to call, what to say, and when to follow up next. It mines insights across demographic, psychographic, geographic and firmographic dimensions by deploying data science across 65000 users and 2500 customers.
Or the case of a user who provides all his frequent flyer logins to an independent virtual agent. The agent books the best flight depending on what he is trying to optimize (cost, dates, use of specific airlines etc.), and what he has done in the past.
Or a bot that helps a marketing professional with targeted recommendations on where to spend his company’s marketing dollars. The agent analyzes patterns across reams of data, both from the company’s past marketing spend, and similar companies.
Independent agents will not rely on advertising, but will monetize through subscription, lead generation or commission based models.
Independent agents of the future will be marketplaces of today infused with the steroids of AI. Or stand-alone, cloud-based AI applications delivering insights and actions. Or an API on top of an AI cloud responding to multiple systems.
Autonomous Agents
The last category is the most transformational, and also the most difficult to achieve.
An autonomous agent is configured and deployed by the customer, and solves for the customer. Not for anyone else. The agent fetches her data from supplier silos, correlates across them to deliver key insights, and also helps her find the best supplier for any job to be done. The user owns — or co-owns — this data and can share parts of it with other parties as appropriate.
For example, autonomous agents can find the cheapest ride by polling Uber, Lyft, Flywheel etc. and optimize for what you desire (cost, comfort, or speed). They can handshake with platform, supplier and independent agents to solve for the customer.
These agents can correlate your nutrition, exercise and healthcare data to find patterns and insights. Or track your shopping data to recommend the best offers across retailers. Or help you securely share your driving data and get better premiums.
Providing customers full agency flips the current model, cuts middle-men out, reduces wasteful marketing spend and unlocks massive value for both customers and suppliers. But to succeed, a couple of thorny problems need to be resolved.
Suppliers must be willing to provide APIs to your (customer) data. Typically, suppliers seek to build their own apps, and have resisted attempts to provide APIs to customers. Mostly because they do not want to lose control and fear losing customers to competition. When will this change? When there is enough demand from a large number of users. Or better yet, when the supplier truly believes that the upside of stronger customer relationships, transparent demand and better access to new customers trumps the downsides of un-bounding the data.
Successful autonomous agents like Adblock add-ons work primarily in single-user mode. They do not need consent or participation of anyone but the user. To succeed in a larger context, autonomous agents need a vibrant marketplace of motivated suppliers and customers. Distribution is a very real bitch for marketplaces without strong incentives for both sides.
Open data government initiatives, regulation protecting rights of consumers to access their data, and several other trends are pushing certain industries — like utilities and healthcare, and geographies — like UK and Australia, to become more amenable to the idea of un-bounding customer data.
There will be a time in the future when the standard for being “easy to do business with” will mean providing an open API to the customer’s data, and customers demand or expect it as the default. When this happens, it will create much larger possibilities for autonomous agents.
In summary,
It is likely the case that the next big thing is not so much a new model that supplants apps, but an intelligent layer that works on the user’s behalf. Cutting across devices, installed and not-installed apps, and other accessible services on the broader internet. Orchestrating and leveraging all of them to fulfill the explicitly stated and implicitly derived needs of its users.
And, when demand is more transparent, tens of billions of marketing dollars are better spent towards what works and what matters. As users engage more with agents to find suppliers for their jobs to get done, guesswork and speculation (a.k.a the ad-tech industry) will lose their hold on the internet.
And the disembodied intelligence that lives in the ether will slowly and surely infuse itself into every aspect of the lives of consumers and businesses.
And also likely cause unintended consequences as these agents try to do some very human tasks without being human.
That seems a long way from now. But may be a lot closer than it appears.