The Role of AI in Customer Experience Management

.

The Role of AI in Customer Experience Management and Marketing

Artificial intelligence (AI) shows up in three places in customer experience management (CXM):

1. Machine learning (ML) works in the background on data, prediction, and decisioning.
2. Generative AI works in the foreground, producing content.
3. Agents work in between, chaining ML and generative capabilities into multi-step tasks.

Most value still sits in the background. Most of the work, too.

The basic idea of any AI is large data sets + iterative processing algorithms + advanced pattern identification = capabilities that mimic human outputs.

If your data is missing or in bad shape, that equation doesn’t work.

This is the big lesson that a lot of companies are currently learning. Despite the hype around AI and especially agents — much of it valid, some of it questionable — the outputs are only as good as the inputs. 

Businesses that invest in the ‘boring’ stuff like governance, logic models, CRM data cleansing, unified customer records, and team capabilities are the ones that reap the rewards from AI in customer experience. 

Background Processes: Machine Learning

For years, “AI in customer experience” has been contained to ML models used behind the scenes. These ‘under the hood’ models analyse data and identify patterns. Depending on the martech, some models also make predictions.

ML enables marketers to surface insights and monitor trends without deep technical expertise.

• Predictive analytics: Forecasting customer behaviours based on historical patterns and intent signals.
• Recommendation engines (Next Best Experience): Identifying, compiling, and delivering the most relevant experience for an individual customer.
• Propensity and risk models: Moderating the risk of churn, over-communication, irrelevant offers, or mistimed communication based on governance logic.
• Adaptive modelling: Continuously self-updating based on observations, to refine recommendations and insights.

While these capabilities were a massive leap forward, they have always operated in the background. A lot of marketers didn’t even realise they were AI (and whether they truly are is debatable) until the last couple of years.

Foreground Creative Tasks: Generative AI Is Everywhere

We all know about generative AI (GenAI) at this point. These are the language models (LLMs) like OpenAI’s ChatGPT and Anthropic’s Claude that work by predicting the next most likely token in a sequence. That might be a word in a document, a pixel in an image, a note in a song or a figure in a spreadsheet.

Of course, this is a massive oversimplification of how GenAI works. But by the time I’ve finished writing this article, one of the big LLMs would have rolled out a new feature — so we’ll keep it simple.

The same goes for how marketers use GenAI. According to HubSpot, 80% of marketers use AI in content creation. I’d wager that’s a low estimate, given the range of tasks it can support:

• Efficiency and cost reduction: Delegating time-consuming busywork like data summaries and early-stage research allows marketers to focus on more valuable work.
• Idea generation: From seasonal campaign to blog topics to offline marketing ideas and internal company newsletters, generative AI can help to get the creative ideas flowing.
• Personalised marketing at scale: GenAI can use pre-approved offer and messaging modules to create personalised, on-brand, in-the-moment customer experiences.
• Automated customer service: AI chatbots can handle customer enquiries 24/7 with almost no capacity limit (although the results tend to be a mixed bag).
• Data analysis: LLMs are getting better at interpreting data sets to find trends, errors, and insights, though I wouldn’t encourage anyone to blindly trust the results just yet.

As GenAI improves, it’s likely to lessen marketers’ admin workloads even more. 

Still, there’s a reason that ChatGPT hasn’t taken your job yet.

For one thing, an LLM can’t think for itself or create something entirely new. What looks like an original text or image is really just a remix of all the data it’s been trained on.

Secondly, customers don’t necessarily AI to take over. A recent PwC survey found that 71% of US consumers prefer interacting with a human than AI, and 75% of global consumers want more human interaction in the future.

Joined-Up Action: AI Agents

An agent is a model given a goal, a set of tools and permission to act. Instead of returning an answer, it plans the steps and carries the task through, calling other systems as it goes. In CXM, that means the work between the insight and the outcome.

• Journey orchestration: Deciding and executing the next step in a customer’s journey based on logic set by the marketer.
• Data preparation: Cleaning, matching, and updating records, including mining ‘messy’ semi-structured and unstructured data.
• Campaign QA: Checking versions, links and segment logic before a send goes out.
• Service resolution: Handling a simple enquiry end to end or escalating to a human after initial triage.
• Internal workflow: Speeding up approvals, asset handling, and briefing to remove process bottlenecks.

Gartner expects 40% of enterprise applications to embed task-specific agents by the end of 2026, up from ~5% in 2025. However, Gartner also expects 40%+ of agentic AI projects to be cancelled by the end of 2027. Unclear business value, escalating costs, and weak risk controls are more likely than poor models to be the culprit. Forrester’s take is similar; agents fail on ambiguity and miscoordination, not bugs.

There’s a vendor problem too. Gartner calls it “agent washing”, estimating last year that only 130 of the thousands of vendors spruiking agentic capability could actually deliver it.

How AI Helps Marketers Solve 6 Common Business Challenges

1. Create a Single Customer View

AI can stitch together fragmented data to de-anonymise visitors. This helps you understand customer behaviour and build comprehensive profiles. Although some de-anonymisation tools might cause GDPR headaches, others focus on customer value.

• Zeta Marketing Platform uses 12.7B digital identifiers and 2.5k+ real-time signals to identify high-value customers and optimise their experience in real time.
• Adobe Real-Time CDP (Customer Data Platform) leverages AI to normalise known and pseudonymous customer data collected from online and offline touchpoints. 
• Apteco FastStats uses AI to segment audiences, find lookalikes, and predict buying behaviour. Our extensive experience with FastStats has helped our clients connect multiple data sources to gain actionable customer insights. 

As well as creating holistic and progressive profiles, these capabilities help you maintain a clean database.

2. Personalise Marketing for Better Results

Customer decisioning AI data models analyse large volumes of data to detect patterns. These patterns can predict which channel, content, product, or message will resonate most with each customer.

This is incredibly valuable for delivering personalisation at scale and in real time. AI tools can then dynamically tailor email content, website experiences, and even app interfaces for each user. 

Most marketing automation platforms (at least the good ones) feature these capabilities to some extent. They tend to veer slightly in one direction, so it’s important to clarify your use case first. For example: 

• BrazeAI™ is popular among B2C brands with a strong mobile presence
• Bloomreach leverages AI to enable e-commerce brands to personalise shopping experiences
• Adobe Target uses AI for A/B testing, personalisation and automation at the enterprise scale
• HubSpot offers AI features across content, sales, customer agent and social media functions

3. Make ad budgets work harder 

AI is foundational to Google Ads. For many years, it has been quietly helping in the background, supporting advertisers in maximising their time and return on investment.

Google

Machine learning has long been pivotal in bidding for and serving online advertisements. AI evaluates campaign performance in real-time, ensuring every dollar delivers maximum impact. Virtually every platform uses this ‘programmatic’ advertising model.

But the one to watch is Bing. Many of the emerging GenAI tools (Copilots, Gemini, Arc Search and others) are currently tied to Bing, Microsoft’s long-suffering Google search competitor. Microsoft Advertising is already rolling out features like conversational ads and chatbot APIs that could help you engage audiences in a different way.

4. Track and Measure Cross-Channel Customer Journeys

Advanced analytics enables marketers to track and measure complex customer journeys with new granularity. AI-powered reporting and attribution modelling help to surface insights and guide decision-making. Essentially, AI marketing tools shortcut data analysis workflows to:

• Map touchpoints from initial awareness to final conversion.
• Surface actionable insights faster from large data sets.
• Measure the impact of marketing efforts on conversions, behaviour and revenue.
• Identify bottlenecks and optimise channel performance.
• Measure the value of multi-channel campaigns more accurately.

Adobe’s Attribution AI feature is a great example of the difference between regular data processing and AI. Attribution AI quantifies the impact of passive and active touchpoints, so you can optimise campaigns to generate the most value. For smaller and growing brands, Tableau is often a more cost-effective BI and data visualisation solution. 

5. Diagnosing Issues and Anomalies

AI can identify unusual patterns in marketing data, such as sudden drops in traffic or conversion rates. In fairness, you can probably do that, too, by looking at reports. Where AI helps is by pinpointing the underlying causes of issues and anomalies so you can take corrective action. Adobe’s Contribution Analysis is a good example. Available in Adobe Analytics, Contribution Analysis is a machine learning process designed to map the contributions to observed anomalies. Most web analytics platforms offer some version of error or anomaly detection. But without the data-processing capabilities of ML algorithms, diagnosing and fixing the issue is often a matter of trial and error.

6. Accelerate Business Processes

AI-powered automation is improving speed and accuracy in asset handling, approvals, and task management. Lots of vendors are rolling out generative tools to help you speed up daily tasks. But back-end automations are just as important for efficient work. This is where agentic AI can benefit brands. Agents can automate the in-between tasks that used to cause bottlenecks. But they need a clear task to be effective. Otherwise they risk becoming a flashy solution with little to no ROI.

The Limitations of AI in Customer Experience Management

AI tools are definitely impressive. And they undoubtedly surpass human capabilities in terms of speed, scale, and volume. But are they actually “intelligent”? Most experts would say no. Don’t worry – we’re not digressing into philosophy. The point is that all AI tools are designed and trained for a specific task.

ChatGPT might seem capable of adaptive thinking. But (oversimplification warning) it’s fundamentally designed to predict the next best word or decision. AI agents might seem like they’re making decisions. But in their current form, they’re not capable of independent thought.

Similarly, Adobe Customer Journey Analytics can predict your customers’ next move faster and more accurately than you could ever hope to. But it’s not trained on website content. It can lead them through the steps that are most likely to land them on your website, but it can’t write the content, interpret their inquiry, or design a unique experience. 

The computer scientist (and co-creator of VR technology) Jaron Lanier summed it up for The New Yorker:

“The most pragmatic position is to think of AI as a tool, not a creature. It’s easy to attribute intelligence to the new systems; they have a flexibility and unpredictability that we don’t usually associate with computer technology. But this flexibility arises from simple mathematics.”

Jaron Lanier

Getting Started with Generative AI in Customer Experience

We see three levels of adoption for GenAI:

1. Off-the-Shelf Tools ($)

Use an existing foundational model(like ChatGPT) directly by inputting prompts directly into its interface.

Depending on the use case, you can speed up tasks like drafting emails and subject lines, brainstorming ideas, writing job descriptions or conducting broad research.

2. Prompt Engineering ($$)

Leverage existing models by programming your own applications and connecting them via API, providing additional context (data, templates or example outputs) to the output you want to achieve. 

This allows you to use public large language models (LLMs) while protecting your IP and generating more useful outputs. 

3. Customising Models ($$$)

Fine-tune existing foundational models by adding layers of proprietary data to significantly alter how the model works. Given the enormous amounts of data in foundational models, any data you introduce needs significant scale. This makes customisation expensive and potentially prohibitive without a solid long-term business case.

What’s the Catch? The Risks of AI in Customer Experience Management

Accuracy

GenAI still generates incorrect or misleading information. These ‘hallucinations’ can seem plausible until you fact-check the content. Machine learning algorithms, by design, are less prone to hallucinations. However, they can still yield insights that don’t quite fit your use case or misinterpret nuanced data.

Data Privacy and Security

AI relies on vast amounts of data, including sensitive customer information. Data breaches and leaks can have severe consequences for businesses. There are also ethical questions around how AI collects and analyses data. Privacy, consent and transparency are increasingly important for users.

Bias 

AI algorithms can perpetuate biases present in the data they’re trained on. They may inadvertently reinforce stereotypes and prejudices, particularly in areas like advertising and content generation.

Skills Gap

The rapid evolution of AI requires marketers to continually upskill and adapt to new technologies, or risk becoming obsolete. We’re not so concerned about the job losses that some pundits predict. There will always be plenty of work for switched-on marketers. Automating low-level and repetitive tasks is more like an opportunity to upskill or retrain.

Dependency

Excessive reliance on AI can stifle creativity and innovation. AI-generated content also tends to lack the emotional connection and authenticity that human-created content can provide. This is why it’s so important to remember that AI is a tool, not a colleague.  

Evolving Regulation

Article 50 transparency obligations under the EU AI Act (aka Regulation (EU) 2024/1689) took effect in August 2026, with fines up to €15m or 3% of global turnover. High-risk obligations were pushed to December 2027. It’s  likely the first of many laws that will restrict or reshape AI in marketing. For example, the Act says: “Content that is either generated or modified with the help of AI…need to be clearly labelled as AI generated so that users are aware when they come across such content.”

You must make an effort to stay informed about changing regulations around AI use.

To mitigate these risks, it is essential to:

• Set limits for responsible AI use within your teams
• Educate your organisation on the uses and limitations of AI tools
• Use high-quality, unbiased data to train AI models
• Implement robust data privacy and security measures
• Monitor AI systems for bias and take corrective action
• Foster a culture of ethical AI use
• Continuously upskill and reskill employees to adapt to the evolving AI landscape
• Maintain human oversight and control over AI-driven decision-making

The Future of AI in CXM and Marketing

We expect that AI models will become more accessible for marketers in the coming months. Vendors outside the enterprise sphere are racing to roll out AI features. The next phase will see AI tools mature. As they evolve and the capabilities improve, you can expect to see efficiency gains and cost reductions in daily tasks. Specialised tools and features will follow, enabling AI tech vendors to differentiate themselves in a crowded market.

Your Business Goals Matter

AI is going to grow up fast. We recommend revisiting your strategies to see where the evolved versions of today’s AI solutions can add value. Don’t get swept up in the hype when there are bigger opportunities to scale your reach and capabilities using AI tools. This could be your opportunity to get ahead of the competition. While others are playing with GenAI, focus on scaling analytics, collecting first-party data, and building a loyal customer base. In other words, invest in customers first. AI marketing tools come later.