Google's Plan to Replace Your Customers

For decades, Google's business model was clear: users search for services, and businesses paid to be found. An entire industry, SEO, arose around the power of this logic.

But Google's latest services now indicate a different future: users give tasks to the Google assistant, it then searches for agents to complete that task. 

Sounds great as a consumer, but has serious implications for businesses.

In April 2025, Google signalled this shift by releasing the Agent Development Kit (ADK) and the Agent-to-Agent (A2A) protocol. Together, they are intended to realise the world where agents search, compare, transact, and orchestrate on behalf of users. In this new paradigm, your customer isn't a person clicking on a link—it's Google's assistant (aka agent) negotiating on behalf of the customer they represent.

But is this hype? After all, how many of your competitors have an AI agent, or even know what these new protocols are? But the truth is, customers will do whatever is most convenient, Google is already in their pocket and will pass deals to businesses who play the game.

Agentic AI: Its Just Too Convenient...

Imagine you want to renew car insurance and have some complexity to consider; age, specialist vehicle, foreign travel etc. It’s time-consuming. You need to look up your prior insurance, draft the requirement, compare the market. It’s a hassle, and you're not an expert—you just want it handled, reliably and affordably.

Now picture this instead: You ask your Google's AI assistant (aka agent) to handle it.

The assistant chats with you to understand your need, but in truth, it already knows you well. It understands the fine print, finds the best rates from providers, weighs their previous service and negotiates directly with their systems. It shows you three appropriate options, you click one and it handles the paperwork.

You'd think this can happen now, but it actually requires a degree of intelligence and business co-operation not available until now. It happens because the user's assistant has these open source frameworks:

• Discover and transact with other agents via A2A

• Form a temporary team of specialized agents using ADK

• Call booking APIs through MCP

Now turn that perspective around: if you run a service business, how do you ensure your service is even in the running?

You need to show up in the agent ecosystem, and that means three things:

Have an agent (A2A): 

• so you can be discovered, negotiated with, and hired on the fly by other agents. No salesperson needed.

Be usable in a team (ADK): 

• so your agent can work in tandem with customs agents, payment agents, weather agents… and not get left out when Google assistants are building their solutions.

Offer a tool (MCP) : 

• even if you don’t have a full agent yet, you should expose your key capabilities as structured APIs that assistants can use. Otherwise, you are too much hassle to deal with.

Strategic Implications

Service providers today enjoy strong human relationships and defensible niches. But agentic AI rewires the competitive landscape. Here’s how:

Buyer Power Increases

• Agents can instantly compare multiple providers based on cost, their experience of the services in previous transactions, whether the provider is willing to pay Google for the work.

Threat of Substitutes

• Y Combinator now backs start-ups which are 'agent first', agents optimised to serve other agents. Open protocols (A2A, MCP) + free toolkits (ADK) lower the barrier to entry.

Brand Disintermediation?

• How will customers value you services or repeat purchases if they don't encounter your brand? Don't worry, the customer is highly attuned to your brand and reputation, more than ever before. This customer is an AI agent, its money is as good as anyone's. Consider appealing to its way of doing things.

Licensing Must Evolve

• Some business models, such as seat-based pricing, don’t scale when AI agents are the customer. They will have one seat but initiate thousands of micro-transactions. Usage-based, machine-readable licenses will be de rigeur.

Agent‑First Playbook

Look at your business through an “agent‑first” lens, how can you help your new AI customers? Some of these are lessons from the world of API services, which are even more relevant with agentic AI:

Agents want trusted partners

• Affiliate with successful agents, those who can promote your service. Pay a kickback for recommendations. SEO advice will surely catchup with Agentic AI.

Agents want to know who they are talking to

• Be prepared to invest in identity technologies for agents. Prove who you are, and in return, demand to KYC (know your customer). Without protection your customers may actually be competitors or worse.

Agents want to know who they trusted

• You may want to prove a service was (or was not) provided by you. Provenance may be a task for zero knowledge proofs.

Agents want fast workflows for their agent teams

• Offer low latency services at a premium, services which humans would not have previously valued.

Agents want refunds and don't appreciate slow human tickets

• Be prepared to give automated judgements for some refunds. This may be a task for smart contracts. Less worrying if you knew your customer.

Agents know teams get things done faster

• Know Your Niche? Build coalitions with complementary businesses to offer packages of services commonly purchased together. Car insurance may be in this category.

Agents want to keep the regulator happy

• Some sectors are regulated, a human must be in the loop. This is radical, but higher value agents can hire humans, see payman.ai by Visa.

Agents want to minimise their tokens

• Don't send bloat, both your agent and your customers' pay to process every token, in and out.

Be the marketplace's early anchor.

• Agent marketplaces are already in place. Some will develop a niche, a service differentiator. The service which exploits that niche will take 70% of the value. Regardless of being an anchor, consider offering generous freemium or latency tiers to remove friction for agent developers.

Final Word

Agentic AI won’t just change how services are consumed—it will change the dynamics of who your customer is. The next interface is no interface. Build for the agent, and the humans will follow.

I'd expect Microsoft to launch a B2B equivalent of Google's B2C strategy. They have been quick to move on agents, copilots are only their first step. Keep an eye on Microsoft Build conference, due May 19-22 2025.

This article is dedicated to those embarking on new businesses and those curating growth. Bon chance!