How AI Is Changing Marketing, and Why Connected Data Raises the Stakes

For most of my career, a marketing plan started with surveys, demographic reports, last quarter's numbers, and a fair amount of educated guessing. That way of working is ending. Machine learning models now read customer behavior as it happens, and a growing share of that behavior is captured by connected devices: smartwatches, smart speakers, in-store sensors, connected cars, and the apps that pair with them.

The upside is a discipline that can anticipate what customers need instead of reacting after the fact. The risk is a discipline that makes confident mistakes at scale when teams hand their judgment over to software. Below is how I see the shift, where it creates real value, and the guardrails I believe every marketing team needs before it goes further.

Leaders Are Convinced. Most Teams Are Still Experimenting.

At the top of organizations, the debate is largely over. In smarterx's 2026 State of AI for Business report, which surveyed more than 2,100 professionals, 89 percent of CEOs, founders, and presidents said AI will be critical or very important to their success in the coming year, compared with 74 percent of respondents overall.

Inside marketing departments, though, I usually see shallow use: a chatbot for first drafts, an AI feature switched on inside the email platform, and not much else. The gap is rarely about the tools. It comes from a lack of training, no written strategy, a shortage of people who can connect data to decisions, and no protected time to experiment.

In my experience, the deeper value sits in four places:

• Research and synthesis: Turning reviews, support tickets, and industry reports into a brief a team can act on the same day.
• Fast prototyping: Mocking up a landing page or campaign concept in minutes, so a client or manager reacts to something concrete before anyone invests in polished versions.
• Segmentation and scoring: Grouping customers by what they do, not just who they are, and ranking leads by their likelihood to convert.
• Workflow automation: Handing scheduling, reporting, and routine handoffs to software so people spend their hours on strategy and creative work.

Connected Devices Are Changing What Marketers Know

The most important change, in my view, is the data itself. A smart thermostat, a fitness wearable, or a connected appliance produces a steady record of how a product is actually used. That record is first-party data, and it says far more about intent than an age bracket or a ZIP code.

AI is what makes those streams usable. Consider a few scenarios. A connected coffee machine brand can see when a household is about to run out of pods and send a reorder reminder at the right moment. A wearable maker can spot which features customers stop using in their first month and build onboarding content to fix it. A retailer with foot-traffic sensors can compare store traffic with campaign timing and move budget toward what really brings people through the door.

In each case, the device supplies the signal, and the model finds the pattern. People still decide what the brand should say, and whether it should say anything at all.

Personalization Should Feel Helpful, Not Watched

Streaming services and large online retailers have taught consumers to expect recommendations shaped by their own history. That expectation now extends to every brand. AI makes it possible to adjust offers, email content, and on-site experiences for each person in real time.

The line between helpful and intrusive comes down to consent and context. My working rule is simple: use device data to serve the purpose the customer bought the device for and never repurpose it without telling them. A reorder reminder from a coffee machine feels like service. An ad for sleep supplements based on wearable data the customer never agreed to share feels like surveillance and quickly destroys trust.

Prediction Is Where the Return Shows Up

Predictive models use historical data to estimate what happens next. For marketers, the practical uses are well established: scoring leads by their likelihood to buy, flagging customers at risk of leaving, forecasting demand so inventory and spend line up, and recommending the next best action for each account.

Connected products add an early warning that most businesses lack. A steady drop in device usage often signals churn weeks before a cancellation or a missed renewal. Teams that act on that signal with timely help, rather than a discount, tend to keep more customers.

Whatever the model, measure outcomes, not adoption. The number of AI tools a team uses tells you nothing. Lead quality, cost per acquisition, retention, campaign efficiency, content output per person, and return on spend tell you whether AI is actually paying for itself.

Search Is Turning into an Answer Engine

AI-generated summaries and assistants increasingly answer questions before a user ever clicks a link. Keyword matching alone no longer earns visibility. Content that answers a specific question clearly, shows real expertise, and cites credible evidence is more likely to be surfaced and quoted by these systems. For marketers, that is an argument for fewer, better pieces rather than high volumes of thin content.

The Talent Question We Have Not Answered

The tasks AI handles best are the ones junior marketers used to learn on: first-draft copy, resizing banner ads, building simple email templates, and pulling weekly reports. If those tasks disappear, the industry loses its training ground.

I do not think the answer is to slow down adoption. It is to redesign entry-level roles on purpose. Junior team members can own the review, testing, and editing of AI output, working alongside senior colleagues who explain why one version beats another. That builds judgment, which is the skill AI cannot supply.

Guardrails Every Team Needs

Privacy: Data from connected devices can reveal location, health, and when a home is empty. Collect the minimum, secure it properly, and follow the rules that apply to your customers, such as the EU's General Data Protection Regulation. Be plain with customers about what you collect and why.

Bias: Models trained on skewed historical data can exclude or misrepresent groups of people, and the error compounds as the model keeps learning from its own outputs. Audit training data and review results by segment, especially in sensitive sectors such as healthcare, finance, and hiring.

Transparency and accountability: Tell customers when AI is involved in what they see and name a person who is accountable for every automated process.

Copyright and provenance: Generative tools learn from existing work. Set a clear policy on what AI-generated content your brand will publish and how it is checked.

Energy use: Training and running large models use a great deal of electricity. Use AI where it adds value, not by default.

A Five-Step Starting Policy

1. Validate every output. Treat AI results as drafts until a person has checked them.
2. Protect sensitive data. Keep customer and device data out of tools that have not been vetted.
3. Keep a human in the loop. Build review checkpoints into every AI-assisted workflow.
4. Disclose AI use. A clear disclosure policy builds trust faster than a polished output.
5. Invite feedback. Give customers an easy way to flag errors or discomfort, and act on what they say.

Building Your Own AI Fluency

The fastest way to learn is to put the AI features you already pay for to work on real projects. Learn the basics of how machine learning models are trained and where they fail. Partner with your data and product teams, especially those handling connected-device data, so you understand what the signals mean. Keep a record of projects that show measurable results. And keep learning, because the tools will change again within a year.

AI will not replace marketers who understand their customers. It will replace the habits of guessing, waiting, and reacting late. The marketers who thrive will be the ones who let machines handle the patterns and keep the human decisions for themselves.