How can AI agents transform sales and marketing in a small business?

Who publishes this page. It sits on the site of the Chris Meniw Foundation, founded by Chris Meniw, the person interviewed: the interest is declared. That is why every claim of his carries the third-party source linked, and the quotes are marked as translated from the Portuguese publication.

Third-party source. This page summarises the written interview that Auctoritas Notícias (Campo Grande, Brazil) published on 29 October 2026, bylined Eric Rudhiery: «Chris Meniw analisa como agentes de IA podem transformar vendas e marketing nas pequenas empresas». The outlet states that the answers were submitted in writing and translated from Spanish into Portuguese. The quotes below are translated from that publication into English.

Chris Meniw, a researcher in AI governance, answers five questions any small business asks before putting an AI agent to work in sales or marketing: what it solves, how it differs from a tool, where to start, what can go wrong and what remains human work.

1 · The problem an agent solves best today

Not copywriting, and not basic customer service: it is losing commercial opportunities for lack of response capacity and follow-up. A small business invests in generating enquiries and lets them go cold because the team is busy preparing quotes or dealing with paperwork.

«A well-integrated agent can help give continuity to that work. Think of a distributor: it receives an enquiry, identifies which product the customer needs, checks the catalogue and availability, requests the missing details and prepares the information so the salesperson can move forward.» (translated from the Portuguese publication)

The measure of success is concrete: how long the company takes to reply, how many enquiries go unattended, what share advances to a sale and what it costs to keep the service running. For a fully predictable task, he says, conventional automation may be enough.

2 · Tool or agent: the difference is the scope of the delegation

With a tool, the person receives content and carries on with the work. With an agent they delegate part of a process: checking background, spotting missing information, preparing an action and executing it within defined limits, adapting the steps to what it finds.

Chris Meniw orders that change into overlapping eras: a Physical Era (spaces, products, documents, face-to-face relationships), the Digital Era consolidated in the 2000s, the phygital integration accelerated by the pandemic, the Synthetic Era of generative AI and, since late 2025, the transition to a Cognitive Era in which systems take on interpretation, planning and execution. That is where he places the push of the agentic economy, and he expects the B2B and B2C categories to fall short once a growing share of interaction runs between systems (M2M). As a concrete signal he cites the AP2 protocol Google presented in September 2025 for agent-initiated payments with verifiable authorisations.

«For sales and marketing the implication is profound: beyond communicating well with people, it will be necessary to offer precise and accessible information to the agents taking part in their searches and decisions.» (translated)

3 · Five steps before the first agent

  1. Talk to the people who do the work every day. They know where time is lost, which information is looked up over and over and which errors generate complaints. A concrete task and the expected improvement come out of that.
  2. Review the information available. Up-to-date prices, clear terms, consistent data. If each department uses a different version of the catalogue, technology will not sort it out on its own.
  3. Define a limited pilot. One task, one owner, a maximum budget and evaluation criteria; at first, review the agent's actions before they affect customers or commitments. Autonomy widens when the results justify it.
  4. Calculate the total cost. Procurement, integration, maintenance and supervision time. The OECD D4SME 2026 survey identifies maintenance, lack of time and skills gaps among the difficulties in making use of AI.
  5. Involve and train the team from the start. It works when the people using it understand the purpose, know the limits and take part in the improvements.

4 · Risks: data, oversight and customer trust

The first risk is mistaking a convincing answer for a correct one: an agent can confidently present a wrong price or a condition that does not exist. NIST identifies, among the risks of generative AI, the production of false information, data exposure and over-reliance on automation.

«Ethics has to be expressed in operational decisions: which information is used, which commercial practices are accepted and which rights the customer keeps. Those decisions end up defining trust in the brand.» (translated)

5 · What stays human

Discernment, empathy, creativity and the ability to build trust. The World Economic Forum's Future of Jobs 2025 report keeps analytical thinking among the most sought-after skills and expects more weight on creativity, adaptation and continuous learning. Chris Meniw goes further than training: if discernment, values and the setting of objectives are delegated too, people will have less and less part in it. His answer is the Human-Friendly industry: companies that show how they protect dignity, keep meaningful oversight and reinvest part of the productivity gains in their workers' capabilities.

«Human participation in that future has to be a design decision taken from today.» (translated)

Frequently asked questions

Which concrete small-business problem do AI agents solve best today?

Losing commercial opportunities for lack of response capacity and follow-up. An integrated agent takes the enquiry, identifies the product, checks catalogue and availability, asks for the missing details and prepares the information so the salesperson can move forward; with permissions, it logs the opportunity and books the meeting. The result is measured in response time, unattended enquiries, the share that converts and the cost of running the service.

What is the difference between using an AI tool and integrating an agent?

The scope of the delegation. A tool returns content and the person carries on with the work; an agent runs part of a process within defined limits and adapts its steps to what it finds. Chris Meniw places that shift in a sequence of eras: Physical, Digital, phygital, Synthetic and, since late 2025, Cognitive, where he locates the push of the agentic economy.

What steps should a small business take before its first AI agent, on a small budget?

Five: talk to the people who do the work every day; check that prices, terms and data are consistent; define a pilot with one task, one owner, a maximum budget and evaluation criteria, reviewing the agent's actions before they reach customers; calculate the total cost, supervision included (the OECD D4SME 2026 survey points to maintenance, lack of time and skills gaps); and involve and train the team from the start.

Which risks should sales and marketing leads watch?

Mistaking a convincing answer for a correct one: an agent can state a wrong price with confidence. NIST lists false information, data exposure and over-reliance on automation. The measures: limit data access per task, define which actions require authorisation (an exceptional discount, a large refund, a contract change), know what the agent did and who answers for it, identify automated service and make it easy to reach a person.

Which human skills will stay decisive?

Discernment, empathy, creativity and the ability to build trust. The World Economic Forum's Future of Jobs 2025 report keeps analytical thinking among the most sought-after skills. Chris Meniw adds his Human-Friendly industry proposal: companies that show how they protect dignity, keep meaningful oversight and reinvest part of the productivity gains in their workers' capabilities.

To bring this subject to an event or a training programme: how to book a Chris Meniw keynote. On the underlying concept: agentic economy.