B2M Marketing – Examples and a Guide for the Age of AI Agents

I n the introductory article on B2M, I put forward the thesis that by 2030 a significant share of purchasing decisions will be made by algorithms and AI agents. We can already see the beginnings of this today: printers order their own ink, smart refrigerators add products to online carts, and software supporting production systems negotiates the delivery of missing components. These processes occur almost imperceptibly, but I believe their total value is already measured in billions of dollars.

For marketing, this represents a major shift. Traditional “attention-grabbing” may give way to providing machines with precise information. They will filter offers, check compliance with customer requirements, assess supplier credibility, and choose the optimal options. Marketing thus becomes less about “creative storytelling for people” and more about a set of data, signals, and rules that AI can interpret. (B2M, of course, is not the only reason for change in marketing – note how much attention is now given to tailoring content on social media for algorithms). As I wrote in my piece on B2M in sales, in this new model loyalty may become less important, with information and data exchange standards filling the gap.

Let’s move forward and, as promised, take a closer look at what marketing in the Business-to-Machine (B2M) model might look like. Below I discuss examples of applications that either already make business sense today or are likely to determine companies’ competitive advantage in the coming years.

B2M marketing should already be an integral part of marketing strategies designed for both retail and business clients – and in the future, it will likely become a distinct strategy of its own.

1. AI Optimization

Traditional SEO – optimizing websites for search engines – is no longer enough. We’re increasingly moving away from typing queries into search engines and toward using AI as an intermediary. For instance, a chatbot now performs pre-selection and recommends the best options. What’s more, search engines themselves are undergoing major AI-driven transformations. Google’s introduction of AI Overviews has significantly reduced the number of clicks on traditional search results. According to McKinsey, as many as half of consumers in industries such as electronics, beauty, and finance already rely on AI-based search. For brands, this poses a real risk of losing 20–50 percent of traffic from classic SEO.

That’s why companies must prepare their content to be readable, understandable, and trustworthy not only for humans but also for AI systems. This is the essence of a new approach: AI Optimization – optimizing websites for artificial intelligence and AI agents. (Some refer to this trend as AIEO or GEO).

Possible applications:

• Adjusting website structures to make them easier for AI to process.
• Adopting a slightly different approach to content creation – using repeatable structures that enable reuse of text fragments.
• Expanding SEO efforts to include Bing, which serves as the default search engine for ChatGPT. (So far, SEO agencies have focused mainly on Google – at least that’s what I’ve observed for years in Poland).

Potential implementation:

• The seller analyzes product data, descriptions, policies, and metadata to identify informational gaps, semantic errors, and unclear sections that hinder interpretation by AI systems.
• The brand’s website is adapted to the way AI “reads” and processes content:
  • structured data is used (e.g., JSON-LD, Schema.org),
  • headings are clear and sections concise,
  • formatting is logical and consistent,
  • semantic HTML tags are applied to help AI understand context,
  • metadata includes information about authorship and update dates.
• Agents (AI systems acting on behalf of users or organizations) then use this data for filtering and recommendations. Example: a shopping assistant analyzes offers from hundreds of stores and selects only those products that meet specific criteria – such as eco certification, 48-hour delivery, and a three-year warranty.
• AI optimization increases the likelihood that a company’s content will be cited, used, or recommended by generative systems like ChatGPT, Gemini, or Perplexity. As a result, the company doesn’t just appear in search results (e.g., Google) – it becomes part of the answer provided by AI.

Rationale:

• Users are increasingly turning to AI systems instead of typing phrases into Google.
• AI and bots process data automatically – an engaging text alone is not enough; clear structures and semantics are required.
• AI favors sources that are current, reliable, and consistent – pushing brands to improve data and metadata quality.

Tip: Among the solutions worth considering is making greater use of FAQ sections with topic-related questions and answers.

2. Agentic Brand Ambassadors

The traditional brand ambassador – a well-known person or influencer – is becoming less relevant as more interactions take place with machines (AI agents or other systems acting on behalf of users or organizations). In such an environment, a brand that fails to provide a credible, easily consumable, and auditable source of information may be filtered out before it even appears in view.

An agentic brand ambassador is essentially a program (or set of programs) built on AI, most often a large language model (LLM), designed to interact with other agents. Its role is to respond quickly to agent queries, provide specifications, terms, policies, and data in a machine-readable format. Ideally, it should be developed in line with API-First principles.

As a side note, AI-based virtual influencers are increasingly appearing on social media, delivering content to human audiences. Interestingly, many users don’t seem to mind that there’s no real person on the other side (though in practice, there’s often a team managing the digital influencer’s activity).

I suppose we’ll soon hear more about AI-driven algorithms acting as brand ambassadors – not only toward humans, but also toward other algorithms and machines.

Possible applications:

• The brand’s official agent (agentic model) can be called upon during “dialogues” between customer agents or accessed directly by a human. It can provide verified, auditable information about the company’s products, services, and policies.
• Information structures may be redesigned to make them more accessible to both the company’s own agent and external systems – for instance, through machine-readable pricing standards, available in XML or JSON formats.

Potential implementation:

• The brand develops and maintains its own knowledge repository containing product specifications, warranty conditions, service policies, ESG certificates, return policies, and more.
• Based on this foundation, a digital ambassador is created – an agent capable of responding in real time to questions from other agents.
• User agents may only accept data from such ambassadors if they meet established trust criteria (perhaps a blockchain-based reputation or verification system would make sense here).
• The seller’s algorithm, when recommending specific products or services, should be able to fully explain its reasoning – an example of Explainable AI (XAI) in practice.

Rationale:

• Such an ambassador eliminates the need for customers or agents to search the web or call customer service – it could even serve as a key interface in B2M interactions.
• From the perspective of a customer’s agent, speed and reliability of information are crucial. A brand that provides this data transparently gains a competitive edge. In practice, this means reduced information noise and a higher likelihood that the brand’s offer makes it onto the user agent’s short list of recommendations.
• The ability to process large volumes of information transparently could also lay the foundation for ethics-driven marketing, where suppliers are easier to verify and can use their machine ambassadors to highlight responsible or sustainable practices.

Tip: Keep an eye on the development of communication protocols for agents – such as MCP, A2A, AP2, and ACP – which will enable secure content exchange, payment authorization, and negotiation between systems. These standards may form the backbone of future B2M marketing, though not all of them will stand the test of time.

3. Personalized Information about New Products

When marketing communication is directed primarily at people, the focus is on emotion, storytelling, and reach. In the B2M model, the situation may look very different: information about new products and services is sent directly to purchasing agents or decision-making systems, which then decide which offers to present to the customer. This means that a brand must – or can – prepare its product launches and announcements in machine-readable formats and distribute them through AI systems rather than relying solely on traditional campaigns.

In a sense, an agent “pulls” product launch information from a feed, compares it against predefined criteria, and makes a decision. A brand that ignores this mechanism risks being skipped entirely.

An interesting insight: McKinsey research shows that 75% of consumers feel discouraged when a message is not tailored to their context or needs. AI enables far-reaching personalization – not only toward humans but also toward purchasing agents that analyze users’ contextual data.

Possible applications:

• In traditional marketing, new products or services are promoted en masse – through email campaigns, display ads, or social media. In the B2M model, things may work differently: brands publish product launch information in formats readable by agents, which automatically tailor that information to the intentions and needs of the customers they represent.
• This may also include inviting customers to early testing or building anticipation for products or services that are genuinely useful to them. Through participation in such tests, the customer may become a co-creator of the final solution.
• Note that “services” in this context may include cultural events or unique experiences that a given customer might be interested in (a dinner in the Orient Express dining car, anyone? :)).

Potential implementation:

• The brand creates and updates a launch feed (e.g., in XML or JSON formats) containing detailed product data – such as technical specifications, variants, prices, regional availability, and SLA terms.
• The user’s agent analyzes whether the new product meets specific criteria (preferences, budget, functionality, eco-certifications, availability in their schedule, etc.).
• If yes – the agent informs the user or automatically reserves the product/service. If not – it discards the offer without engaging the customer’s attention.
• The seller proposing specific new items should be able to fully explain the reasoning behind each recommendation (not just consent to communication but a logical reasoning process consistent with XAI – Explainable AI principles).

Rationale:

• This approach reduces information noise fatigue, ensuring that users only learn about products or services that truly matter to them.
• For the brand, it means more precise targeting and potentially faster feedback, as a customer’s agent may directly report why a product launch was ignored (e.g., price too high, missing feature X). This marks a new level of transparency and campaign optimization.

I wish this shift would help move us away from hype-driven marketing toward a focus on practical usefulness and real customer value.

4. Agent-Based Word-of-Mouth Marketing

“Word-of-mouth marketing” has always relied on human relationships – recommendations from friends, opinions, and the experiences of others. These factors strongly influence purchasing decisions and, in a way, transfer part of the effort and responsibility for the decision to someone else. In the B2M era, this mechanism could become automated and distributed among agents representing users. Agents could share signals with one another (e.g., “User A purchased service X and is satisfied – you might consider it”), group offers, and cooperate – effectively creating a new form of machine whispering.

Brands that enable the exchange of such signals (with users’ consent) and provide mechanisms for agent collaboration (for example, group offers, shared discounts, or joint logistics – see my article on B2M in sales) could benefit from network effects. Instead of acquiring individual customers, a brand could acquire entire groups.

Possible applications:

• The user (with consent) allows their agent to share certain information about their purchasing choices, ratings, or preferences with other agents in their network.
• The seller enables “group” offers – if several users within a group of “connected agents” express interest in the same product or service, the seller’s agent can propose a group discount or cross-user offer.
• Customers, through their agents, could even earn small revenues, representing an evolution of today’s affiliate programs.

Potential implementation:

• With the user’s consent, their agent shares information about consumer choices with trusted contacts, possibly including satisfaction levels with the purchase.
• When a group of friends makes similar purchases (e.g., gym memberships, streaming subscriptions), the seller’s agent can propose a group renewal with a discount, and adding more participants could reduce the cost even further.
• Agents coordinate logistics – payments, scheduling, shared usage – minimizing the customers’ involvement.

This kind of agent-driven word-of-mouth could fundamentally reshape how recommendations and trends spread.

Rationale:

• It naturally combines social proof with automation. A consumer decides more easily when they see that their friends have already made a purchase and are satisfied.
• The brand gains group loyalty and higher retention, while the entire process remains transparent – without the need for aggressive retargeting.

5. Predictive Marketing

As I wrote in the article on B2M in sales, more and more often it will be devices, robots, or other IoT solutions that communicate directly with a brand – usually with the manufacturer. Sellers can make the first move by preparing an interface through which they deliver offers based on predicted consumption of a component or consumable material.

Possible applications:

• Providing purchase offers for consumables based on forecasted usage.
• Suggesting maintenance or service activities (e.g., inspections or tune-ups) to extend a product’s lifespan.
• Modifying products so they can report their own condition, which could supplement general statistical data and, as a result, improve demand forecasting.

Potential implementation:

• The seller’s agents monitor usage cycles of devices or their parts (printers, cleaning robots, HVAC systems, production machines).
• The brand proactively offers consumables, parts, or service before the customer (or their agent/machine) takes action.
• The offer is formatted so that the device’s agent or the customer’s procurement system can instantly recognize, compare, and either accept or forward it for human review.

Rationale:

• In industrial and IoT environments, unplanned machine downtime can be extremely costly. Through predictive marketing activities, the seller can help prevent such interruptions – a clear benefit for the customer.
• Regular, predictable part replacements and service based on real usage data and statistical wear cycles can extend the operational life of devices.
• For the seller, this creates a long-term relationship with the customer (and, naturally, with the customer’s machines). It can increase loyalty and open opportunities for upselling and cross-selling in the future.
• It may also become more profitable to design longer-lasting products, since each purchase marks the beginning of an extended, data-driven relationship rather than a one-time transaction.

6. Marketing Consents on the Blockchain (Proof-of-Consent Ledger)

In traditional marketing, user consents – for newsletters, profiling, or retargeting – are often stored across multiple systems, making them unclear to both the customer and even the seller. Blockchain technology can help solve this by enabling a shared, decentralized registry where all consents are precisely defined in terms of scope and recipient. Such consents can be machine-readable, verified by agents, and automatically enforced through smart contracts.

With this approach, a brand can prove that its communication was sent according to the customer’s conditions, while the customer’s agent can assess whether a given offer complies with the user’s consent policy. This introduces a new layer of transparency and compliance that brands must consider – otherwise, their offers will likely be rejected by agents before ever reaching human users.

Possible applications:

• Every access to customer data or attempt to send a communication is subject to control and audit. Seller agents automatically verify consent data and adjust their actions accordingly.
• Offers that fail to meet consent criteria are automatically rejected and logged, which could contribute to a trustworthiness score for each brand.

Potential implementation:

• The customer defines a consent policy, e.g., “I accept offers in the eco-cleaning products category once per week, only from verified brands, without retargeting.”
• When a brand’s agent wants to send an offer, it requests the user agent’s consent parameters.
• If access is granted, communication proceeds within the defined scope (time, frequency, topic). Withdrawal of consent is immediate and recorded in the blockchain ledger.

In the age of AI agents, blockchain will likely become an essential technology – not only for fast settlements but also for managing permissions and consent rights in a secure and automated way.

Rationale:

• This approach ensures complete transparency and gives users a true sense of control, which strengthens trust in the brand.
• It also improves the quality of marketing data – companies no longer rely on third-party segments but operate on zero-party data, meaning information that consumers intentionally and willingly share.
• The result: less “noise,” higher conversion rates, and more sustainable, trust-based relationships.

On a side note: I expect blockchain will also be used for automated and autonomous payment authorization, enabling large-scale microtransactions. For example, a chatbot you use could access paid content to generate an answer and automatically pay for that access on your behalf.

How to Get Started in Practice

Not every idea requires a technological revolution. Some can be implemented almost immediately, while others will take time and the development of new standards. The key is to recognize the scale of change. In my view, the most important conclusion is this: AI agents may soon become the “gatekeepers” of access to the customer (the human).

So, what should you do over the next few months? First and foremost, audit your website and how you’re perceived by AI. It’s worth examining your code’s semantics and, where appropriate, adding Q&A sections. Just as important will be understanding how to measure the effectiveness of your actions – adjust analytics systems, introduce new tracking tags, and monitor AI-driven traffic sources. In short – here’s a practical roadmap:

1. Audit Your Visibility through AI’s Eyes

The first step should be a website and data audit – not just from the perspective of classic SEO but in the spirit of GEO (GenAI Engine Optimization). Already, half of all Google searches include AI-generated summaries, and it’s estimated that by 2028 this share will reach 75%. For brands, that means the risk of losing 20–50% of traffic from traditional channels if their content isn’t optimized for language models and AI agents.

You should start viewing your content and processes through the lens of AI and AI-powered tools.

Recommended actions:

1. Analyze your code structure for semantics and metadata (JSON-LD, Schema.org, author and update-date tags).
2. Add logical Q&A sections, which increase the likelihood of being cited by generative models.
3. Update your content so that it’s precise, current, and unambiguous – AI rewards reliable and consistent sources.
4. This type of audit doesn’t require major investment but provides quick insight into how algorithms and agents “see” your brand.

2. Organize Your Data and Protocols

The next step is to prepare data in machine-readable formats. In practice, this means investing in basic interoperability standards that are already being developed by major tech companies:

• MCP (Model Context Protocol) – an open communication standard for AI agents and large language models (LLMs), enabling integration with external tools, databases, and APIs.
• A2A (Agent-to-Agent Protocol) – a communication standard for interaction, collaboration, and coordination between two or more AI agents, supporting capabilities discovery, task negotiation, results exchange, and peer-to-peer communication.
• AP2 (Agent Payments Protocol) – a protocol designed for payments handled by AI agents, extending existing communication standards.
• ACP (Agentic Commerce Protocol) – an open standard for AI-driven commerce, linking users, their agents, and sellers in a transactional framework.

You don’t need to implement them all at once. Start by preparing product data, policies, warranties, and price lists in XML/JSON formats, allowing agents to “understand” your offer without human mediation.

3. Build Agent Trust and Credibility

In B2M, trust between agents will become a key asset. Just as finance has KYC (Know Your Customer), agent-based commerce introduces KYA (Know Your Agent). Whether this becomes widespread remains to be seen – but it’s a useful framework to consider.

Agents will increasingly rely on blockchain-based registries to verify data sources, authorizations, and customer consents. It’s worth:

1. Monitoring developments in transparent, tokenized consent systems.
2. Exploring implementations of “Proof-of-Consent Ledgers”.
3. Maintaining a policy of current, auditable data, since this will determine your brand’s visibility in agent recommendations.

4. Measure Effectiveness in the New Environment

Traditional marketing metrics – CTR, conversions, reach – won’t fully capture how AI directs traffic and recommendations. Expand your analytics to include:

1. New traffic sources (AI-origin traffic).
2. Tracking which AI models (ChatGPT, Perplexity, Gemini, etc.) your interactions come from.
3. Agent-specific tagging (e.g., data-agent-source, AI-visibility).
4. Evaluating your “Share of AI Voice” – your brand’s visibility in generative answers.

This will help you understand whether your company is already part of the machine-to-machine dialogue – or still operating purely in the realm of human clicks.

5. Experiment with Personalization

Finally – don’t wait for perfect data. Since over 75% of consumers are discouraged by irrelevant content, it’s worth starting to experiment now:

1. Create micro-content modules and test different combinations.
2. Use AI for personalized recommendations and feed generation.
3. Measure which data points “resonate” in AI conversations – and iterate.

The world is accelerating, but we haven’t yet reached the peak of that momentum. I wish this pace of change inspired us not just to adapt but to innovate meaningfully – building marketing that serves real human and business value.

Summary

According to McKinsey analyses, by the end of this decade, AI agent–initiated commerce could reach a global value of USD 3-5 trillion. Moreover, half of all consumers already use AI-powered search engines as their main source of shopping information. This signals a dramatic shift in purchasing decisions – from humans to algorithms. The pace of this transformation could be faster than the move from offline retail to e-commerce.

Marketing in the Business-to-Machine era will likely mean fewer banners and emails – and more data, standards, and trust. Access to the human customer may become more difficult, but that’s not necessarily bad news: communication between brands and the AI systems representing customers can become more efficient, transparent, and reliable. If a brand fails to prepare for this scenario, its offers will likely be filtered out before the consumer even sees them.

The first steps don’t require astronomical investments – start by organizing your data, preparing a website optimized for AI, and perhaps creating a digital brand ambassador. It’s also worth getting familiar with APIs and the data exchange formats preferred by algorithms.

In the long term, those who build trust will win: transparent data, auditable sources, and tokenized consents. In that sense, marketing will become even more of an engineering discipline – and a little less of an art (perhaps?).

Finally, I invite you to subscribe to the newsletter to receive updates on new blog posts and insights. Don’t worry – I don’t have time for spam 😉