Executives & Directors · Module 05 · Chapter 005

AI in Sales and Marketing

Sellers report spending most of their week not selling, and AI may give much of that time back. The same AI makes the commercial message almost free and targeting precise, which turns two old questions into executive decisions: which customer signals are fair to use, and what counts as success. Point AI at the scarce conversation, build every claim from approved truth, and judge it by qualified pipeline and retention, not volume.

≈ 16 min read

After this chapter you can

Map where AI helps across the commercial cycle, and redesign a seller's week so the seller keeps judgment and commitments.
Distinguish a lead score or forecast from a seller's qualification.
Explain why AI sales coaching must be fitted to the seller's level of performance.
Require commercial content to be generated from approved sources, including claims about AI itself.
Treat customer signals, personalized prices and AI disclosure as leadership decisions with legal anchors.

In the spring of 2024, Salesforce surveyed 5,500 sales professionals in 27 countries. Sales reps said they spent 70 percent of their time on tasks other than selling: research, preparation, administration, data entry. Four in five sales teams were already experimenting with AI or had implemented it. Yet only 35 percent of sales professionals completely trusted the accuracy of their own data¹.

Figure 5.5.1 The time is there to win back. The data that AI would learn from is trusted by barely a third.

The survey is a vendor’s and the figures are self-reported, but they describe something every sales leader recognizes. The first promise of AI in sales is obvious: give the seller back the hours. The second promise is less obvious and more dangerous. AI also makes the commercial message almost free to produce and the target almost free to find. When those two costs collapse, the limits on a sales and marketing organization stop being capacity and start being judgment: whom to contact, with what claim, using which signals, and measured how.

A commercial cycle, not a content factory

A common first move with AI in marketing is to produce more: more emails, more campaign variants, more social posts. It is rarely the most valuable. Content is one stop on a longer cycle. Marketing creates and shapes demand. Sales turns demand into pipeline and pipeline into revenue. Account teams keep and grow what was won. What happened feeds the next round of decisions.

Figure 5.5.2 AI can help at every stage. The question at each one is whether a better commercial decision came out.

AI can help at each stage. It can read signals that already exist in CRM records, web visits, product usage, call notes and campaign responses. It can estimate who is likely to buy or leave. It can draft the proposal, the email and the follow-up. It can summarize a call and suggest the next step. The test is the same everywhere: did a better decision come out about who, when, what and which offer, or only a thicker customer profile and a larger send list?

How to count the money that results, and how much of it truly belongs to AI, is the subject of Where AI Creates Revenue. This chapter stays with the work itself and with the decisions that only leaders can make about it.

A seller’s week, before and after

Start where the Salesforce numbers point: the hours around the conversation. A typical complex sale involves account research before the first call, preparation before each meeting, notes and CRM updates after it, a proposal, and a chain of follow-ups. Each of these is language work that today’s AI can draft. That is a capability: independent evidence that it gives sellers measurable time back is still thin, and most published figures come from vendors.

Figure 5.5.3 An illustrative week. AI takes the drafting and the recall; the seller keeps the judgment and the commitment.

The right-hand column matters more than the middle one. The redesign is not “the AI does the admin”. It is a new division of labor in which the machine drafts and recalls while the seller decides and commits. Leaders who skip that column get notes nobody checks and follow-ups nobody would have sent.

There is also a decision about what the recovered hours are for. If nobody decides, they drift into more email. As Where AI Creates Revenue argues, saved selling time is capacity, not revenue, until someone turns it into more and better customer conversations.

A score ranks; a seller qualifies

Prediction is where AI changes who gets the seller’s attention. A lead-scoring model estimates how likely each lead is to convert and what it might be worth, then ranks the queue. In effect it says: this lead resembles leads that converted before. That is useful, because seller time is the scarcest input in the function.

Figure 5.5.4 The score sets the order of the queue. The seller decides what deserves to stay in it.

A score is not qualification. A seller still has to establish whether there is a real need, whether timing and budget are real, and whether the opportunity fits the strategy. Forecasts work the same way. CRM reporting tells you what is in the pipeline; an AI forecast estimates what is likely to happen. A sales leader should always be able to ask why the forecast moved: deal stage, engagement, timing, historical pattern. A forecast without drivers is hard to trust and easy to ignore.

Two cautions belong here. First, a model knows only its history. If only a third of sales professionals fully trust their data, a model trained on it inherits the doubt, along with old territories and old habits. Second, the most likely buyer is not always the best target. Eva Ascarza’s field experiments found that customers at highest risk of leaving were not the ones whom a retention offer helped most²; Where AI Creates Revenue develops the point. A next-best-action suggestion is only as good as its reason. “Call the customer” is weak. “Call, because usage rose and the renewal window opens next month” is something a seller can weigh. How forecasts become decisions is the subject of AI in Prediction, Forecasting and Optimization, later in this module.

AI coaching helps the middle of the team most

AI does more than rank leads. It can listen to calls and coach the seller. Some of the strongest evidence on what happens next comes from a set of randomized field experiments with sales agents, published in the Journal of Marketing in 2021. Xueming Luo and colleagues compared agents coached by AI with agents coached by human managers³.

Figure 5.5.5 The gain from an AI coach followed an inverted U. An AI coach paired with a human manager beat either alone.

The benefit followed an inverted U. Middle-ranked agents improved the most. Bottom-ranked agents were swamped by the volume of feedback the AI produced; when the researchers restricted it to fewer points, their performance improved. Top-ranked agents showed the strongest aversion to being coached by a machine. A combination of the AI coach and a human manager outperformed either one alone³.

For a leader, the lesson is not “buy a coaching tool”. It is that one design does not fit the whole team. The newest sellers need less information, not more. The best sellers need a reason to trust the tool, and usually a manager who uses it with them. Rolling out the same assistant, with the same settings, to everyone wastes most of its value.

Generate from approved truth

Sellers spend hours on proposals, and marketers on copy. AI can combine the customer’s requirements with product information and earlier documents into a first draft in minutes. The critical control is what the draft is built from.

Figure 5.5.6 Commercial content is generated from approved truth, checked against it, and approved by a person wherever it commits.

Drafting from approved sources, rather than from whatever the model absorbed in training, is the retrieval pattern described in RAG and Enterprise Knowledge — Executive Mental Model⁴. In sales and marketing it needs one more piece: a claims library. Product capabilities, prices, delivery terms, statistics and competitive comparisons should come from a list someone owns and keeps current. Brand voice and visual identity form the rest of the fence. Inside it, a small team can produce at scale. Outside it, cheap content turns into unsupported statistics, outdated features and promises nobody approved.

The message is now almost free

Here is the picture to carry. AI has made the commercial message almost free to produce. When the cost of something collapses, people use far more of it.

Figure 5.5.7 Cheap messages tempt volume. Volume without relevance buys noise, and noise costs trust.

The channel itself already enforces this. Since 2024, Gmail has required anyone sending more than 5,000 messages a day to its users to authenticate their mail and, from June that year, to offer one-click unsubscribe, processed within two days⁵. Senders must keep the spam rate their recipients report below 0.3 percent, and Google recommends staying under 0.1 percent⁶. A campaign that triples volume and annoys one recipient in three hundred is no longer only a brand problem. It risks not being delivered at all.

So the useful question for an AI campaign is not how many messages it can produce, but how few it needs. Sometimes the right strategy is fewer, better interactions, measured against a group that did not receive them. Running that comparison honestly is the discipline of controlled experiments⁷.

Signals, prices and disclosure

Precise targeting raises the question that volume does not: which signals are fair to use? The test developed in AI and Customer Experience applies here: would the customer reasonably expect this use of their information? In sales and marketing the question has four concrete forms, and the first is what the company says about AI. US regulators have already penalized firms for overstating their use of AI. When the FTC launched Operation AI Comply in 2024, its chair put it plainly: there is “no AI exemption from the laws on the books”⁸; the securities cases are told in What Exactly Is Artificial Intelligence?⁹. The FTC’s 2024 rule on reviews and testimonials goes further for marketers: it bans fake reviews, including AI-generated ones, and lets the agency seek civil penalties¹⁰. An “AI-powered” claim in a campaign needs the same substantiation as any other claim, and a review written by a model is not a customer’s review.

The second is purpose. Data a customer gave for one reason, such as a service request or a health record, is not automatically available for a campaign. The third is price. AI makes it possible to set a different price for each customer, and at least one US state now requires the business to say so. The fourth is disclosure. Customers react when they learn they were dealing with a machine. In one field experiment, telling customers up front that a sales call came from a bot cut purchases sharply, even though the bot sold as well as skilled staff¹¹; Where AI Creates Revenue covers the study. The answer is to build assistants good enough to be disclosed, not to hide them.

The lawRules that already apply to commercial AI (as of October 2026)

US federal, claims about AI: there is no AI-specific federal statute, but existing law already applies to what companies say about AI. Claims about AI in marketing, investor materials and product descriptions need the same substantiation as any other claim (SEC antifraud rules; FTC Act Section 5). Regulators have already penalized firms that overstated their use or capability of AI to investors (SEC) and to consumers (FTC).
US federal, reviews: the FTC Rule on the Use of Consumer Reviews and Testimonials (16 CFR Part 465, in effect since 21 October 2024) bans creating, buying or disseminating fake or false reviews and testimonials, including AI-generated ones and reviews by people with no real experience of the product. The FTC can seek civil penalties per violation.
New York, personalized prices: a business that sets a price for a New York consumer with an algorithm that uses that consumer’s personal data must show a clear and conspicuous notice: “THIS PRICE WAS SET BY AN ALGORITHM USING YOUR PERSONAL DATA”. It is a disclosure duty, not a ban on personalized pricing. In effect since 10 November 2025, after a federal court dismissed a First Amendment challenge; an appeal was being considered¹².
EU, AI Act Article 50: since 2 August 2026, transparency duties apply: disclose AI interaction, and label synthetic and deepfake content; generative systems already on the market before then have until 2 December 2026 for machine-readable marking.
EU, GDPR: personal data needs a lawful basis and is limited to the purpose for which it was collected (Arts. 5-6).

Financial, health and other sensitive data carry extra protection in most jurisdictions. Under the GDPR, health data is special-category data: processing it is prohibited unless a specific exception applies (Art. 9). This is orientation, not legal advice. Confirm with counsel.

Autonomy raises the stakes on all three. An agent that picks accounts, writes the message and sends it is making targeting, claims and disclosure decisions at machine speed. Start with an approved segment, approved messages, a frequency cap and a person who approves the send, then widen the limits as results come in. How to set those limits is the subject of AI Agents and Intelligent Workflows.

Story: the tax return and the ad tag

Online tax-preparation services compete for customers in a short season each year, and they advertise heavily. The big ad platforms run AI that decides who sees each ad, and that AI learns fastest when it is told who converted. The platforms offer advertisers a small tracking tag for their websites. Put it on the pages where customers act, and the platform’s AI learns who signs up and files, and finds more people like them.

Before reading the record, take the decision yourself.

Figure 5.5.8 Three proposals. Decide before reading on.

Option A promises the best-trained targeting. Option B gives up digital marketing. Option C keeps the tag off any page that reveals a customer’s finances, measures sign-ups and filings with the firm’s own analytics, and tests campaigns against holdout regions.

Now the record. In November 2022, The Markup reported that several large US online tax-filing services, among them TaxAct, H&R Block and TaxSlayer, had the Meta Pixel on their sites. Depending on the service, it sent Facebook’s parent company details such as income, filing status, refund amounts and the names of dependents. The firms used such tags to target advertising, and most removed or changed them after the reporters asked¹³.

Figure 5.5.9 The remedy was not a better tag. It was a decision process, written down and audited.

In July 2023, an investigation by seven members of Congress reported that the tags had exposed data on tens of millions of taxpayers over at least two years, and that Meta used such information to target advertising and to train its algorithms. The lawmakers asked federal agencies to investigate. Meta answered that advertisers should not send it sensitive information and that it filters such data; TaxSlayer said the report contained false or misleading statements [@warren-tax-privacy-report-2023; @cbs-tax-prep-meta-2023].

The costs then arrived in installments. TaxAct settled a class action in December 2024. In August 2026, Connecticut’s attorney general announced a further settlement: TaxAct, now under new ownership, would pay 275,000 over data shared with Meta and Google from January 2018 to December 2022. It must also set up a review committee for third-party tracking, write policies for tracking technology, document the data points each tag collects, scan its website with a tag-monitoring system and commission two independent audits¹⁴.

Look at what the remedy is. It is not a better tag. It is a decision process: someone with authority decides which signals may leave the company, writes the decision down, and checks that the website still matches it. Option C is where that process starts. The lesson reaches well beyond tax. When you feed your customers’ signals into someone else’s AI, you have made a data-sharing decision, not a marketing-technology setting. That decision belongs to leadership, before the tag goes live.

What this means for leaders

AI in sales and marketing pays first in the hours around the conversation, then in better targeting, and only when it is bounded does it pay in revenue that lasts. Four decisions stay with you. Decide what recovered seller time is for. Fit the assistance to the seller, because one design does not fit the whole team. Own the claims library and the signal list, because the AI will use whatever it is given. And ask for outcomes, measured against a comparison group, before anyone reports volume.

Check yourself

A lead score tells sales who will buy.
In field experiments, an AI sales coach helped middle-ranked agents the most.
If an AI tool wrote the marketing claim, the company is not responsible for it.
Gmail requires large senders to keep reported spam below 0.3 percent.
New York bans prices set by algorithms using personal data.
Adding an ad platform’s tracking tag is a technical setting, not a data-sharing decision.

Reflection: trace one signal

What comes next

Sales and marketing make the promise. Customer service is where the promise is kept or broken, at scale and every day. The next chapter, AI in Customer Service, looks at the difference between answering a customer and resolving the problem.

Laws referenced

Not legal advice. Laws change; verify before relying on this, and consult counsel for decisions.

EU AI Act · EU

Regulation (EU) 2024/1689, as amended by Regulation (EU) 2026/1744

Risk-based rules. Prohibited practices include social scoring, untargeted scraping of facial images, and emotion recognition in workplaces and schools (with narrow exceptions). High-risk systems (Annex III: biometrics, safety components of critical infrastructure such as energy, water and traffic, employment and worker management, credit, education, essential services, law enforcement, migration, justice) need risk management, data governance, documentation, logging, human oversight, human oversight that keeps people able to understand the system, notice automation bias (over-reliance on its output), override it or stop it (Art. 14(4)), appropriate accuracy, robustness and cybersecurity (Art. 15), automatic logging of events (Art. 12), a provider quality-management system (Art. 17) and conformity assessment. An Annex III system is not high-risk if it poses no significant risk of harm, for example a narrow procedural or preparatory task that does not replace human assessment; systems that profile people are always high-risk, and a provider relying on this exception must document it and register (Art. 6(3)). Deployers of high-risk AI must use it as instructed, assign competent human oversight, monitor its operation, keep logs for at least six months and report serious incidents (Art. 26); employers must inform workers' representatives (Art. 26(7)). Public bodies, private providers of public services, and deployers of credit-scoring or life and health insurance pricing systems must carry out a fundamental-rights impact assessment before first use (Art. 27). Providers must run post-market monitoring (Art. 72). A deployer that puts its name on a high-risk system, substantially modifies it, or changes its purpose so that it becomes high-risk takes on the provider's obligations (Art. 25(1)). A substantial modification (Art. 3(23)) of a high-risk system needs a new conformity assessment, unless the change was pre-determined and documented at the first assessment, as with planned continuous learning (Art. 43(4)). Providers of general-purpose AI models (from 2 Aug 2025) must keep technical documentation, have a policy to comply with EU copyright law including text-and-data-mining opt-outs, and publish a sufficiently detailed summary of training content (Art. 53). Research, testing and development before a system is placed on the market or put into service is outside the Act, except testing in real-world conditions (Art. 2(8)). Since the 2026 Omnibus, the Art. 4 AI-literacy duty is an obligation of effort (take measures to support literacy), not of result. Fines reach EUR 35 million or 7% of global turnover for prohibited practices.

2024-08-01 — Entered into force
2025-02-02 — Prohibited practices (Art. 5) and the AI-literacy duty (Art. 4) apply
2026-07-27 — Omnibus softens Art. 4: providers and deployers must take measures to support AI literacy; no specific level must be guaranteed
2025-08-02 — General-purpose AI model obligations apply; governance and penalties regime in place
2026-08-02 — Transparency duties (Art. 50) apply: disclose AI interaction, label synthetic and deepfake content (marking for generative systems already on the market: 2 Dec 2026)
2027-12-02 — High-risk obligations for Annex III systems (e.g. hiring, credit, education, essential services) - moved from 2 Aug 2026 by the 2026 Omnibus
2028-08-02 — High-risk obligations for AI in products regulated under Annex I

Last verified 2026-10-06 · official text

General Data Protection Regulation · EU

Regulation (EU) 2016/679

Personal data is any information relating to an identified or identifiable person, directly or indirectly, including by an identifier such as an online ID (Art. 4(1)). Lawful basis and purpose limitation (Arts. 5-6); processing special-category data, including biometric data used to identify a person, health data and data revealing ethnicity, is prohibited unless a specific exception applies (Art. 9); data protection by design and by default (Art. 25); processors such as AI vendors may act only under a written contract with required terms and sufficient guarantees (Art. 28); transparency to data subjects (Arts. 13-14); right not to be subject to a decision based solely on automated processing with legal or similarly significant effects (Art. 22); breach notification to the supervisory authority within 72 hours (Art. 33) and to individuals without undue delay when the risk is high (Art. 34); data protection impact assessment for high-risk processing (Art. 35). Fines up to EUR 20 million or 4% of global turnover.

2018-05-25 — Applies

Last verified 2026-10-08 · official text

US enforcement against false AI claims ("AI washing") · US - federal

Federal securities antifraud rules and Investment Advisers Act Marketing Rule (SEC); FTC Act Section 5 (unfair or deceptive practices)

There is no AI-specific federal statute, but existing law already applies to what companies say about AI. Regulators have penalized firms that overstated their use or capability of AI to investors (SEC) and to consumers (FTC). Claims about AI in marketing, investor materials and product descriptions need the same substantiation as any other claim.

2024-03-18 — SEC's first AI-washing cases: Delphia and Global Predictions settle for USD 400,000 in total civil penalties
2024-09-25 — FTC launches Operation AI Comply, with five actions over deceptive AI claims and uses

Last verified 2026-10-08

FTC Rule on the Use of Consumer Reviews and Testimonials · US - federal

16 CFR Part 465

Bans creating, buying or disseminating fake or false reviews and testimonials, including AI-generated reviews and reviews by people with no real experience of the product; bans incentives conditioned on positive reviews, undisclosed insider reviews and fake social-media influence indicators. The FTC can seek civil penalties per violation.

2024-08-14 — Final rule announced
2024-10-21 — In effect