The Next Customer Won’t Be Human

In my last post, The Next Employee Won’t Be Human, I wrote about what happens when companies can add capability and capacity without adding equivalent human headcount, and why that changes the economics of building new businesses.

AI agents won't just change who does the work. They are also going to change who buys things.

For most of economic history, companies have designed products, distribution channels, payment systems and customer experiences around a fairly obvious assumption: somewhere in the transaction there is a human making a decision. We search, compare, negotiate, approve and eventually pay.

AI agents are starting to insert themselves into each of those steps, initially helping us make better decisions but increasingly taking action on our behalf. An agent might research a product today, negotiate a purchase tomorrow and eventually manage an entire category of spending within a budget and set of permissions without a human approving every transaction.

That creates an interesting question for companies thinking about where future growth comes from:

What happens when machines become customers?

Most of today's enterprise AI conversation still revolves around copilots, where a human asks a question, the AI produces an answer and the human ultimately decides what happens next.

Agents change that relationship because they can increasingly receive an objective, determine the steps required, interact with other systems and execute work with varying levels of autonomy.

The more interesting examples aren't necessarily personal travel assistants booking hotels. Think about a large company's procurement agent continuously evaluating thousands of suppliers based on price, availability, performance and contractual terms, or a small business agent managing software, insurance, banking and other services within parameters established by the owner. A new venture could have agents continuously testing advertising channels, purchasing services, managing cloud infrastructure and reallocating spending based on what is actually working.

In each case, the human establishes the objective, permissions and constraints while the agent increasingly handles the decisions and transactions underneath them.

That is an important transition because AI moves from influencing economic decisions to making economic decisions.

Once that happens, agents don't simply become better software.

They become economic participants.

Most of our digital infrastructure assumes the user is a person. Websites are designed for human eyes, checkout flows are designed for humans to click through, identity systems authenticate people and financial accounts ultimately assume there is a human somewhere approving what happens.

We have literally spent the last twenty years making humans prove they aren't robots, which may turn out to be slightly awkward.

The next generation of commerce could increasingly involve software agents interacting directly with other software agents. A corporate procurement agent could negotiate with a supplier's pricing agent, an insurance agent could continuously evaluate coverage against changing business requirements, or a venture's operating agent could automatically purchase services and allocate resources as the company grows.

At that point, the competitive advantage isn't necessarily having the best website or the smoothest checkout flow. It becomes whether your products, pricing, inventory, reputation and terms can be discovered, understood and transacted with by machines.

That is not simply a technology upgrade.

It creates an entirely new surface for growth innovation.

This is where the intersection between AI and the future of finance becomes particularly interesting to me.

If an agent can research, negotiate and make decisions but can't independently exchange value, its economic autonomy remains limited. Agents will eventually need some combination of identity, permissions, budgets and payment infrastructure that allows them to act while keeping humans and companies ultimately in control.

A company probably doesn't want its marketing agent to have unrestricted access to the corporate bank account. It might be perfectly comfortable, however, giving that agent authority to spend up to $25,000 per month across approved platforms, reallocating that capital based on customer acquisition performance within predefined parameters.

This starts to look less like giving a machine a corporate credit card and more like programmable capital, where the rules and permissions increasingly travel with the money.

That creates potential new growth opportunities for banks, payment companies, fintechs and other financial institutions because agent identity, authorization, fraud prevention, payments, treasury and auditability all need infrastructure.

Agentic finance starts connecting intelligence directly to capital.

This is also why I've become increasingly interested in the convergence of AI with stablecoins, tokenized assets, smart contracts and programmable wallets.

Individually, each technology has interesting applications. Together, they start looking like potential infrastructure for an economy where machines increasingly participate alongside humans.

Stablecoins provide digitally native money that can move globally and continuously, smart contracts allow rules to be embedded into transactions, programmable wallets can give agents tightly controlled authority over capital, and tokenized assets make financial products accessible through software-native infrastructure.

That doesn't mean every agent needs a crypto wallet, and traditional banks, payment networks and fintechs will almost certainly build substantial agentic capabilities into existing financial rails.

The broader point is more important: financial infrastructure increasingly needs to become machine readable, machine accessible and machine executable.

For incumbent financial institutions, that should create a much more interesting growth innovation question than how to add another AI assistant to the existing banking app.

What financial products need to exist when the customer accessing them is increasingly software?

The implications extend well beyond financial services because companies have spent decades optimizing how they acquire and serve human customers.

We build websites, optimize conversion funnels, buy search advertising, create loyalty programs and obsess over reducing the number of clicks between discovery and purchase.

An AI agent doesn't care whether your homepage has a beautiful hero image or whether the button is the perfect shade of blue. It cares whether it can understand what you're selling, determine the price, evaluate the terms, establish trust and complete a transaction.

For the last twenty years, companies have optimized for Google. The next decade may require them to increasingly optimize for agents.

That creates potential new businesses around machine-readable product infrastructure, agent marketplaces, identity, trust, reputation, pricing, APIs and machine-to-machine distribution. Companies that already own valuable data, customer relationships, distribution or trusted brands may have significant advantages, provided they don't assume the next generation of customers will behave like the last one.

The customer experience of the future may increasingly be an agent experience.

This is where I think the opportunity becomes particularly relevant for CEOs, boards and capital allocators.

The obvious response to AI is to use it to improve the existing customer experience, reduce service costs or make current products easier to use. Those are worthwhile investments, but they are still largely about optimizing or transforming the core.

The bigger growth innovation opportunity is asking what entirely new products, services and businesses become possible when agents can discover, negotiate, purchase and transact autonomously.

A bank might build the financial operating layer for autonomous businesses. An insurer might create coverage that is dynamically priced and purchased by agents as risk changes. A marketplace might be designed primarily for machine buyers and sellers. An industrial company could expose spare capacity or services directly to procurement agents. A software company might build an entirely new distribution model where agents discover, trial, purchase and configure its products without a traditional sales process.

Many large companies already own the assets required to pursue these opportunities: customers, proprietary data, industry expertise, distribution, capital and trust.

The challenge is that these opportunities probably won't emerge naturally from the existing product roadmap.

This comes back to the argument I've made throughout this series. Companies need dedicated entrepreneurial teams operating outside the core that can combine corporate assets with AI-native capabilities, test new business models quickly and progressively allocate capital as evidence builds.

The objective isn't simply to use AI to run today's business better.

It's to discover the businesses AI makes possible.

The internet dramatically reduced the cost of distributing information, cloud computing reduced the cost of accessing computing power, and AI is now dramatically reducing the cost of intelligence.

The next shift is connecting that intelligence to economic action.

When agents can research, reason, decide and transact, they stop being simply another layer of software and start becoming participants in the economy. That means companies shouldn't only be asking how AI changes their workforce or makes today's customer experience better. They should be following the agent further downstream and asking what happens when it starts making decisions, allocating budgets, selecting suppliers, buying products and interacting directly with other agents.

That is where I think some of the most interesting new growth opportunities will emerge.

If you want to understand what to build next, follow the agent. Follow where it needs identity and trust. Follow where it needs data and permissions. Follow where it needs to discover and evaluate products. Follow where it needs to move money. Follow where existing infrastructure assumes a human is sitting on the other side of the transaction.

Every point of friction is potentially a new product, platform or business.

Large companies already own many of the assets required to build into these opportunities, including customers, proprietary data, distribution, industry expertise, trust and capital. The strategic question is whether they use AI primarily to make the existing business more efficient or allocate some of that capital toward discovering the businesses that an agentic economy makes possible.

We are still early, and there will be significant questions around security, fraud, accountability, regulation and how much autonomy people and companies are ultimately willing to delegate. Those constraints matter, but they also create opportunities for the companies willing to solve them.

In my last post, I asked what companies could build when the next employee doesn't have to be human.

This time, I think the more important question is: If your next customer might be an AI agent, what should you be building for it today?