I think Voice AI is a transitional technology. And given the amount of money pouring into companies like Decagon, Cresta, Sierra, PolyAI, and others, I recognize that might seem like an unusual position to take right now.
These companies are doing genuinely impressive things. The technology can now handle conversations that would have been extraordinarily difficult to automate just a few years ago. AI can authenticate customers, investigate problems, access enterprise systems, complete transactions, and carry on conversations that increasingly resemble those handled by experienced service employees.
The money behind this market reflects that progress. In January, Decagon announced $250 million in new funding at a $4.5 billion valuation. Gartner reported in August 2026 that AI spending among customer service leaders had increased 38%, while overall service and support budgets had grown just 2%.
At the enterprise level, the financial commitments can become enormous. A multiyear contact center modernization program involving Voice AI, telephony, core platforms, systems integration, cloud infrastructure, implementation, and organizational change can potentially exceed $30 million or $40 million. That represents the possible cost of a broader enterprise program, rather than a typical standalone Voice AI contract, but it illustrates the scale of the decisions companies are making.
I understand the enthusiasm. If your business receives millions of calls annually, and AI can handle a substantial percentage of those conversations more efficiently than human employees, there's a compelling financial opportunity. Customers can get faster service, employees can spend more time on complicated problems, and the company can reduce its operating costs.
But I've been thinking about what happens after this initial wave of investment. What happens when companies become better at identifying and correcting problems before customers contact them? What happens when my personal AI can contact a company's AI, authenticate itself, and resolve an issue without either of us needing to have a conversation?
And there's another question I've become particularly interested in. When my personal AI acts on my behalf and a company's AI represents the business, which one has the better understanding of me? Who controls that understanding, and which one ultimately influences my decisions?
I suspect the answers could reshape customer service, CRM, customer data platforms, personalization, loyalty, and much of the customer experience technology market. We may be investing billions in technology that perfects the contact center shortly before we substantially reduce the need for it.
At the same time, the most valuable customer information may begin moving beyond the direct reach of the brands that have spent decades trying to collect it. Both changes could happen together, and I think they're closely connected.
We've spent decades making it easier for customers to report problems
Consider how customer service generally works today. Your bank declines a legitimate transaction. Your airline cancels a flight. Your internet provider charges you twice. Your insurer increases your premium, or a retailer ships the wrong product.
In most of these situations, something has happened within the company's operations, systems, or processes. Yet the responsibility frequently falls on the customer to discover the problem, determine whom to contact, explain what happened, and wait while somebody attempts to fix it.
We've built an enormous industry around helping customers navigate that process. First came call routing, interactive voice response, and large contact center operations. Then websites, mobile applications, knowledge bases, live chat, and chatbots. Over the past several years, companies have invested in AI that assists employees, recommends answers, summarizes conversations, and predicts why customers might contact them.
Now AI can increasingly conduct the entire conversation. That's meaningful progress, particularly for companies struggling with labor costs, inconsistent service quality, complicated systems, and long resolution times.
But consider the declined banking transaction again. The bank knows that the transaction failed because its systems declined it. It has information about the customer, merchant, transaction, account history, and potentially other signals that could help determine whether the transaction was legitimate.
We can build an exceptionally capable Voice AI agent to handle the customer's subsequent call, investigate the situation, and remove the restriction when appropriate. We can also work toward systems that recognize the likely problem, determine whether a correction is safe and permitted, and request the customer's approval before completing the necessary action.
The second approach requires deeper integration with operational systems, stronger controls, and more coordination across the business. It also creates the possibility that the customer never needs to initiate service. That is a different long-term investment opportunity.
I believe the next decade of customer service will involve three overlapping developments: automating conversations, automating the work required to resolve problems, and preventing some of those problems from requiring customer involvement at all. The third has the potential to produce the greatest improvement in customer experience, although the second is likely to create significant value sooner.
A fourth development may change the commercial relationship between customers and companies even more profoundly. Customers are beginning to bring their own AI into the equation.
Future 1: The contact center becomes largely automated
Let's begin with the future that most enterprises are already preparing for. Over the next several years, I expect AI to handle a growing share of routine voice interactions. Customers will increasingly be able to change reservations, check account information, request refunds, troubleshoot technical problems, manage subscriptions, and complete other common tasks without waiting for a human employee.
The financial opportunity is substantial. Imagine a company handles 10 million customer contacts annually, with an average direct handling cost of $6 per contact. That's $60 million in annual handling costs.
If the company successfully automates four million of those contacts at an average cost of $2 each, the gross handling-cost difference is $16 million annually, before implementation, oversight, escalation, and other expenses. Those are illustrative numbers, and actual savings depend on the work being automated and which costs can genuinely be removed. Still, they demonstrate why companies are investing aggressively.
I expect Voice AI providers to generate substantial revenue in this market, and many enterprises will achieve worthwhile returns from their investments. However, the business case generally begins with an assumption that the company will continue receiving a large volume of customer-initiated interactions.
That assumption could become less reliable. If a company becomes better at preventing service failures, correcting problems automatically, and accepting requests from customers' own agents, the number of traditional conversations could decline.
An enterprise might find itself operating an extremely efficient automated contact center while the amount and nature of customer service work change around it. This is where the long-term risk begins.
Future 2: Companies start fixing problems before customers contact them
Imagine you're traveling, and your airline cancels your flight. Today, you might receive a notification directing you to an application where you can review alternatives. If the options are unsuitable, you may need to contact the airline and explain your circumstances.
Now imagine the airline detects the disruption, reviews your itinerary, identifies alternative flights, checks your seating preferences, and reserves a suitable replacement. You receive a notification explaining that your original flight was canceled, another itinerary has been held, and the alternative would get you to your destination 45 minutes later. You approve the change, and the airline completes the work.
You never contact customer service. Similar possibilities exist in telecommunications, banking, retail, insurance, utilities, and subscription businesses.
A telecommunications provider could detect and correct a configuration problem before it causes a prolonged outage. A retailer could recognize that a delivery is likely to miss its promised arrival date and arrange an alternative. A bank could identify an account issue, evaluate the available remedies, and seek approval before the problem becomes more disruptive.
Some companies already provide narrower versions of these capabilities. What changes over time is the range of problems they can recognize, how accurately they can determine the appropriate response, and how much work their systems can complete.
I don't expect every problem to become predictable or preventable. Companies won't always know that a product arrived damaged, that a service interaction was frustrating, or that a technically correct decision created a poor experience.
There are also important limits on what companies should do without permission. Changing a financial account, replacing an insurance policy, or accepting a travel alternative can have significant consequences for a customer.
Still, I believe companies will become progressively better at correcting problems without waiting for customers to initiate the process. That has enormous implications for traditional customer service technology.
A system capable of detecting and preventing a problem may create more value than one that handles the resulting phone call. Yet many technology investments are currently justified based on the number of conversations they can automate. I would want enterprises to evaluate both opportunities.
Future 3: Your personal AI becomes your customer service representative
Now imagine a personal AI that understands your subscriptions, purchases, travel plans, household commitments, account relationships, and preferences, based on information you've chosen to make available to it. You tell it to monitor your telecommunications bills and flag unexpected increases.
One morning, it notices that your provider has increased your monthly charge by $12. It checks your current agreement, reviews available plans, and contacts the provider's AI to request an explanation.
The provider's agent verifies the authorization, checks the account, and determines whether a correction or alternative offer is available. The two systems exchange the necessary information, complete an approved change, and document the outcome.
You receive a short summary explaining what happened and what was done. There was no call, chatbot session, or service case for you to manage.
I can imagine this happening across a wide range of services, including subscription cancellations, warranty claims, airline compensation, insurance renewals, billing disputes, and product returns. The technology required to support these interactions is beginning to emerge.
On October 1, 2026, Decagon introduced its Personal Agent Gateway, designed to help businesses serve customers who send their own AI agents to request assistance. On October 6, the company introduced the Personal Agent Consent & Trust Protocol, developed with Instinct, to help companies establish which outside agent is making a request and what the customer has authorized it to do.
These are important developments because identity and authorization become fundamental when software begins acting for people. An agent might be authorized to check an account balance without being permitted to transfer money. It might be allowed to investigate alternative travel arrangements without having permission to cancel an existing reservation.
The company also needs confidence that the agent genuinely represents the customer and that the requested action complies with its own policies. Standards such as the Agent2Agent protocol are intended to make communication between different AI systems easier. But agreeing on a common way to exchange information doesn't automatically resolve questions about customer consent, liability, financial risk, or disputed decisions.
Those questions will take longer to address. Even so, I suspect agent-to-agent service will become increasingly common because it can eliminate a great deal of unnecessary work for customers. And it introduces a question that could be more consequential than the disappearance of the phone call.
Future 4: Two agents, two very different models of the same customer
Let's imagine you're booking a hotel. Marriott's AI has a substantial amount of information about your relationship with the company. It knows your previous stays, loyalty status, room preferences, spending patterns, service complaints, and the offers you've accepted.
It may use that information to predict which properties you're likely to consider, how much you're willing to spend, whether you're likely to return, and which upgrades might appeal to you. Your personal AI potentially knows something quite different.
It has information about your travel across multiple hotel chains, your preferred airlines, your family's schedules, what you've spent on recent trips, and how much you're planning to spend on this particular vacation. It may also know that you want to stay within a certain budget because you'd rather spend the additional money on an experience with your family.
Marriott might reasonably conclude that you're an excellent candidate for an expensive room upgrade. Your personal agent might conclude that the upgrade offers little value for this particular trip. Both models could be working with accurate information.
The important difference is the context available to each. For decades, companies have tried to build a more complete picture of their customers by combining transaction history, service interactions, website behavior, loyalty information, survey responses, and predictive analytics.
Much of the promise of CRM and customer data platforms has been based on the belief that better customer information leads to better decisions about service, marketing, retention, and growth. But what happens when a customer's personal AI has a broader understanding of their priorities than any individual brand could reasonably develop? And what happens when that AI begins making decisions on the customer's behalf?
I believe this creates a fundamental challenge for traditional customer analytics. The brand may know everything about its own historical relationship with the customer. The customer's agent may know what the customer is trying to accomplish next, how they evaluate competing options, and which trade-offs matter most to them.
The most valuable customer decisions may increasingly happen outside the brand's ability to observe them. Consider an insurance company whose predictive model identifies a customer as having a low probability of leaving.
The customer has renewed for years, rarely contacts service, and pays consistently. Based on the insurer's historical data, the relationship appears stable.
Meanwhile, the customer's personal AI notices an 18% premium increase, compares competing policies, and finds an alternative that provides similar coverage for $400 less per year. The customer authorizes the switch.
The insurer's customer model may still see a relatively loyal customer because the behavioral signals on which it relies have not materially changed. The personal AI has already initiated the process of ending the relationship.
This could create a substantial challenge for customer segmentation, propensity models, churn prediction, and personalization. Companies won't necessarily lose access to all customer information. They'll continue to have transactional records, direct interactions, and other information they're permitted to use.
But an increasing share of the customer's consideration and decision-making process could take place elsewhere. And I suspect that's where the competition for customer understanding becomes much more interesting.
Who actually owns the customer model?
I see three possible arrangements emerging over the next decade. In the first, brands retain substantial control over their customer models. Customers may use personal agents, but companies continue managing the information associated with their own relationships, and outside agents receive only what is necessary to complete authorized transactions.
In the second, personal AI providers become the primary keepers of customer context. Companies such as Apple, Google, OpenAI, Meta, or others could maintain persistent models that understand individuals across many different commercial relationships. That would give those platforms considerable influence over purchasing decisions, renewals, provider selection, and service interactions.
In the third, customers maintain some form of portable personal model that different AI providers can access with permission. They can decide which information to share with a brand, which preferences an agent may use, and which actions the agent may take.
I suspect the second arrangement is more likely to develop first because major AI providers already have the infrastructure, consumer reach, and technical capabilities necessary to build persistent customer context. But I think the third is potentially more desirable for customers.
Imagine being able to move your accumulated preferences, decisions, permissions, and relevant history from one personal AI provider to another without having to rebuild everything from scratch. That could create more competition between AI providers and give individuals greater control over the information being used to represent them.
There are significant obstacles, however. The personal AI provider may have its own commercial interests. It may influence recommendations through its business partnerships, advertising relationships, or financial incentives.
The customer's agent might appear to understand their preferences while actually relying on incomplete, outdated, or incorrect information. And just because an agent represents a customer doesn't mean that the customer legally owns every prediction or profile created by the software provider.
This distinction is important. Data ownership isn't a single, universal legal concept. Companies may have rights and obligations involving records they collect, customers may have rights over the use of their personal information, and AI providers may maintain additional information derived from their interactions.
The UK Information Commissioner's Office explains that data portability rights do not necessarily extend to inferred or derived customer profiles. Individuals may still have other rights to access information about themselves, so portability and access should not be treated as the same thing.
So I wouldn't assume the future involves one definitive customer model controlled entirely by one party. My expectation is that the brand, the customer's agent, and potentially other service providers will maintain different models with different purposes and different rights to use information.
The commercial advantage will depend on which model is best suited to the decision being made. And that leads to another important question: What happens when those models disagree?
When the two agents disagree, who gets to decide?
Imagine your bank's AI recommends a premium credit card. The bank has analyzed your transaction history, existing products, account behavior, and previous responses to offers. It predicts that you're likely to accept the card and that the relationship would be profitable.
Your personal agent knows you've recently decided to simplify your finances and reduce the number of accounts you maintain. It declines the offer on your behalf, assuming you've authorized it to do so.
Now imagine the bank proposes a different product with a particularly attractive introductory rate. Your agent compares competing products, evaluates the terms, and determines that an alternative institution has a better offer.
The bank's agent may respond with an improved proposal, while the personal agent continues evaluating the alternatives. We could eventually see agents negotiating with each other, with one representing the company's commercial interests and the other representing the customer's preferences and financial objectives.
This raises questions about fairness, disclosure, and trust. The brand shouldn't automatically receive unrestricted access to everything the customer's agent knows. Personal financial priorities, health information, family circumstances, and other sensitive details may have little relevance to the transaction.
Likewise, the customer's agent shouldn't automatically receive access to a bank's internal profitability calculations, fraud models, or confidential business information. Both parties need appropriate boundaries.
What they do need is a reliable way to establish which facts are relevant, what each agent is authorized to do, what has been offered, and which decisions have been approved. I can imagine an arrangement in which the customer's private model, the brand's private model, and a shared record of authorized transactions operate together.
The customer model contains personal preferences, priorities, and information shared across relationships. The brand model contains company-specific transaction history, product eligibility, business rules, and other information necessary to serve the customer.
The shared record documents what each party agreed to, which permissions were used, and what happened. Neither party needs a complete copy of the other's model for this to work.
The industry has begun developing some of the technical foundations for these interactions. Decagon's PACT protocol, for example, distinguishes the identity of a personal agent from the customer's authority granted to it, and provides a mechanism for recording authorized actions. But determining who controls customer information, how disputes are handled, and which decisions can be reversed will require much more work.
I've written separately about why identity is becoming core infrastructure for AI-powered customer experience. The emergence of independent customer agents makes those identity and authorization questions considerably more important.
Future 5: Personal AI creates an entirely new competitive environment
There's another consequence of personal AI that I think deserves serious consideration. For years, brands have invested in creating direct relationships with customers through mobile applications, loyalty programs, personalized communications, websites, and service experiences.
Those investments assume that customers will continue interacting directly with the brand across a meaningful portion of the relationship. But what happens when customers increasingly delegate those interactions to personal AI?
Imagine asking your AI to manage your family's telecommunications services. It reviews your usage, evaluates reliability, compares available offers, checks contract terms, and recommends the best combination of services. If you've authorized it to proceed, it can potentially make the necessary changes.
The same process could apply to insurance, travel, utilities, banking products, recurring subscriptions, and many retail purchases. In that world, the personal agent becomes an important intermediary between the customer and the companies seeking their business.
The brand may still provide the service, manage the account, fulfill its commitments, and maintain its customer records. But another company's AI may influence which providers the customer considers and which offers they ultimately accept.
There is already an interesting tension emerging here. Google's Universal Commerce Protocol is designed to help AI systems work with retailers and payment providers. Google explicitly emphasizes that merchants remain the seller of record and retain their customer relationships.
That may be true in contractual and operational terms. But if the customer discovers the product, evaluates alternatives, makes a decision, and completes the purchase through Google's AI, Google could have substantial influence over the decision-making process.
I suspect this will become a major concern for brands. Companies may continue owning important transactional records while losing visibility into much of the process through which customers form preferences and make decisions.
And the largest personal AI providers could gain considerable influence over how products and services are compared. That could ultimately change how brands think about marketing, customer acquisition, loyalty, and retention.
Personal AI could also create more service demand
There's a complication in this future that I think many executives will overlook. Today, customers ignore an enormous number of small problems because resolving them takes too much effort.
You might overlook a $7 billing error rather than spend half an hour contacting your provider. You might accept a delayed delivery rather than investigate your eligibility for compensation. You might allow an unused subscription to renew because cancellation requires too much time.
Your personal AI has no comparable patience problem. It can examine every invoice, monitor every service agreement, identify every eligible refund, and pursue claims that would previously have been abandoned.
That means companies could experience a significant increase in service requests even as the number of traditional human conversations declines. There are early vendor observations consistent with this possibility.
In an October 20 article, Decagon described personal agents pursuing issues customers previously ignored, including price matches and forgotten subscriptions. The company reported that some request categories among its customers are already experiencing increased contact volume and handling time. That is a vendor's observation rather than an independent measure of the whole market.
The implications are substantial. Imagine millions of customer agents pursuing legitimate requests that previously would never have reached a company. Now imagine the company's AI responding to those requests while managing its own costs, policies, and commercial interests.
We could see an enormous increase in automated transactions, disputes, and negotiations. There is also the possibility that customer agents repeatedly challenge decisions or pursue concessions after a matter has already been resolved.
Companies will need rules that distinguish legitimate customer rights from repetitive or abusive requests. They'll also need to ensure those rules don't become another way of making it unnecessarily difficult for customers to obtain a fair outcome.
The economics could become very different from the assumptions behind many current contact center automation programs. A company might reduce human-handled conversations by 80% while processing far more automated service transactions than it does today.
That future creates opportunities for software providers capable of managing high volumes of authorized transactions efficiently. It also raises questions about consumer protection, fairness, and whether customers with more capable personal agents receive better treatment than those without them.
What happens to CRM, customer data platforms, and CX technology?
I think the implications for the broader customer experience technology market could be significant. For decades, companies have invested in building a more complete understanding of their customers.
CRM platforms capture relationships, accounts, sales activity, and service history. Customer data platforms bring information together from different systems. Experience management platforms collect feedback, identify trends, and help companies understand what customers are experiencing.
Analytics and personalization systems use that information to predict behavior and determine what the company should do next. Those capabilities will continue to matter.
But the availability of personal AI introduces a new source of customer understanding that the company may not own, control, or even have permission to access. The brand's systems may know what the customer has done historically. The personal agent may have a better understanding of the customer's current priorities, emerging intentions, and available alternatives.
That creates an uncomfortable possibility for enterprise technology vendors. A company could spend tens of millions of dollars building sophisticated customer profiles, predictive models, and personalization capabilities, only to discover that a growing share of customer decisions are being made somewhere else.
The company's historical information will remain useful, but its ability to influence future behavior may decline. Consider the implications for a few technology categories.
CRM platforms will remain important for managing business relationships and transactions, but may need to interact more frequently with external AI agents that influence customer decisions. Customer data platforms may need to work with information selectively shared by personal agents rather than assuming the enterprise can continually expand its proprietary customer view.
Experience management platforms may need to measure outcomes that occur without a conventional customer interaction or survey response. Personalization platforms could increasingly be evaluated on how well they respond to a customer's stated priorities, as communicated through their agent, rather than how accurately they predict those priorities from observed behavior alone.
And Voice AI providers may find that speech becomes one of several ways their systems communicate with customers and outside agents. The commercial boundaries between these categories could continue to shift.
It also makes me wonder whether the traditional ambition of creating a complete, proprietary view of the customer becomes progressively less realistic. Companies will still need accurate, relevant information about their customers. They may simply have to become more comfortable working with customer context they don't permanently control.
The Voice AI companies can already see the transition
I don't believe the leading Voice AI companies are unaware of these developments. Their recent product decisions suggest they're expanding their ambitions beyond the conversation.
Decagon's Personal Agent Gateway directly addresses the emerging need to serve customers through outside AI agents. Its work on permissions and authorization suggests that the company understands how much customer service could eventually happen outside traditional channels.
Cresta is also moving beyond the conversation. On October 13, it introduced Deputy Agent, designed to continue handling work across enterprise systems and teams after a customer interaction has ended.
That is a sensible direction. Many service problems involve work that continues long after a conversation finishes. Insurance claims, technical investigations, financial disputes, and complex account changes can require several departments and multiple systems.
A vendor capable of completing that work has a larger opportunity than one primarily responsible for conducting the conversation. I suspect the strongest Voice AI companies will increasingly position themselves around resolution, workflow execution, customer context, and direct communication with external agents.
But they will encounter substantial competition. Salesforce, ServiceNow, enterprise software companies, industry-specific technology providers, and companies operating core transaction systems all have reasons to pursue the same opportunity.
Many already control the systems that actually execute the work. Why should the contact center AI vendor become the technology responsible for coordinating customer needs across the entire enterprise?
Some may earn that position. Others may find that their capabilities become part of a broader platform operated by another provider. This is one reason I remain cautious about assuming that today's fastest-growing Voice AI companies will necessarily become tomorrow's dominant customer experience platforms.
What happens to the $40 million contact center investment?
Let's return to the enterprise making a major technology decision today. Imagine a company preparing to spend $40 million over five years on a contact center modernization program that includes Voice AI.
The business case assumes millions of incoming customer interactions, a meaningful reduction in human handling costs, and significant productivity improvements. Now imagine that over those five years, the company begins preventing more problems, automating the work required to resolve them, and accepting transactions initiated by external customer agents.
Its traditional phone volume declines substantially. That could be an excellent outcome for customers and the company. But the organization may need a very different set of capabilities than those originally prioritized in the investment.
Instead of concentrating primarily on voice automation, it may need technology capable of identifying problems across operational systems, validating external agent permissions, completing transactions without conversational interfaces, maintaining reliable records, and managing customer information shared across different parties.
How much of the original investment supports that future? How much would need to be replaced or significantly expanded? And if the company reduces its incoming voice volume by 50% or 70%, how much of the original business case still holds?
These are hypothetical scenarios rather than forecasts. But they're exactly the kind of scenarios I'd want to test before signing a major, multiyear contract.
I would ask vendors to demonstrate what their technology can accomplish when no customer conversation occurs. I'd want to understand how the platform interacts with external agents, which data and permissions it requires, and whether it can complete complex work across enterprise systems.
I would also want to understand how customer models are managed. Does the vendor create its own persistent customer profile? How does that information relate to the company's existing CRM and data platforms? What happens if the customer authorizes a personal AI to share information temporarily? Can the company use that information beyond the original interaction, and under what conditions?
If the organization changes technology providers, what information, permissions, records, and capabilities can it take with it? And, importantly, does the commercial model remain attractive as the number of conversations changes?
Per-minute and per-conversation pricing can make sense for the current generation of Voice AI. Those arrangements could become less attractive if more customer service work takes place through structured transactions and long-running automated processes. I would want enough flexibility in the contract to accommodate a changing mix of interactions and automated work.
We may need to rethink how customer experience is measured
These developments also have implications for the way we evaluate customer experience. Imagine a company reports that it reduced incoming customer service calls by 50%.
That sounds like a substantial improvement, but it doesn't tell us what happened to the customer experience. Perhaps the company corrected problems before customers needed to call. Perhaps customers began using personal AI agents instead of contacting the company directly. Or perhaps the business made service difficult to access, and customers abandoned their attempts to get help.
Each situation would produce a similar decline in contact volume, but the outcomes are very different. The distinction becomes even more important when personal agents begin completing transactions on behalf of customers.
If my AI resolves a billing dispute without my involvement, what should the company measure? I may not know the dispute occurred until the agent reports the result. A conventional post-interaction survey would have limited value in that situation.
I think companies will need stronger measures of the underlying customer experience. How frequently do problems occur? How many are corrected before the customer becomes involved? How much work must the customer perform? How often do supposedly resolved problems recur?
When a customer agent takes action, was the request completed correctly, within the approved permissions, and with a clear record of the outcome? And when an agent recommends changing providers, what influenced that decision?
I would also want companies to distinguish between customer preferences directly expressed by a person, preferences inferred by the company's own models, and recommendations generated by an external personal agent. Those are different kinds of information, and treating them as interchangeable could lead to poor decisions.
There will still be a role for satisfaction, effort, and loyalty measures. But the way those measures are collected and interpreted will need to change as more service activity happens without conventional customer interactions.
For customer experience leaders, this could create an opportunity to become much more closely involved in how the company actually operates. Preventing a billing error may require changes to finance systems. Avoiding a failed delivery may involve inventory, fulfillment, and logistics. Resolving an account problem automatically may require product, technology, risk, and service teams to work together.
Customer experience teams have often been asked to measure problems they have little authority to fix. A world in which AI can identify and address those problems may force companies to reconsider who is responsible for the experience and who has the authority to improve it. I think that would be a healthy development for the discipline.
What enterprise buyers should consider now
I've been examining this broader convergence in Be Customer Led's independent CX technology research. Our Native AI Conversational Platforms Built for Resolution report evaluates 16 vendors across three market groups. It looks at what it actually takes for AI to complete service work safely in a production environment, including policy enforcement, integration with enterprise systems, human handoffs, and the limits of current platforms.
Those capabilities matter today, but I believe they'll become even more important as the market evolves beyond conversational automation. Our broader CX Technology Market research covers more than 70 vendors across 12 buying categories. It examines where technology providers overlap, which capabilities companies actually need, and how buyers can avoid treating very different products as though they solve the same problem.
The growing convergence between conversational AI, customer data, analytics, automation, and enterprise systems makes that distinction increasingly important. We've also developed the CX Technology Investment & Architecture Index 2027, which helps enterprises evaluate where to continue investing, which technologies to consolidate or reduce, and when they should buy, build, or combine capabilities.
The question of who controls the customer model adds another dimension to those decisions. Companies need to understand whether they're investing in technology that will remain useful as customer-initiated interactions decline, external agents become more common, and customer context increasingly exists outside their own systems.
For organizations considering major contact center or AI investments, Be Customer Led provides independent advice on technology selection, vendor evaluation, and the financial and operating implications of those decisions. I would encourage buyers to evaluate their current requirements carefully while also testing how those requirements might change over the next five to ten years.
A vendor can offer an excellent product today without necessarily owning the capabilities that will matter most in the future. The question is how much of the investment will remain useful as customers and the technology change.
My prediction: The customer model becomes more valuable than the conversation
I expect Voice AI to remain a substantial and potentially very profitable technology market for years. Enterprises have enormous volumes of existing service activity that can be automated, and the current generation of AI is increasingly capable of handling that work.
I also expect companies to become better at preventing service demand, completing work without customer involvement, and communicating directly with personal agents acting on behalf of customers. The traditional contact center should gradually become a smaller part of the broader customer service operation in many industries.
But I think there's an even bigger change coming. For decades, companies have tried to create a more complete understanding of their customers so they could anticipate their needs, personalize their experiences, influence their purchases, and improve retention.
Now customers may begin carrying their own increasingly sophisticated understanding of themselves, through AI that can compare competing businesses and act on their behalf. Brands will continue maintaining their own customer models, and those models will remain essential for many operational and commercial decisions.
However, personal agents could have greater influence over which companies customers choose, when they decide to leave, and what they expect from the relationships they maintain. That creates a very different competitive environment.
It could make traditional approaches to segmentation, loyalty, churn prediction, and personalization less effective when important decisions occur outside the company's visibility. It could also create opportunities for brands that are transparent about their products, reliable in fulfilling commitments, easy for authorized agents to work with, and consistently competitive on the value they provide.
I don't expect companies to stop competing for customer attention or trying to understand customer preferences. But I do expect them to face increasingly capable software that evaluates their promises and performance from the customer's perspective.
And I suspect the company that controls the conversation will become less important than the parties that control the relevant customer context, the authority to act, and the ability to complete the work. That brings me back to the original Voice AI investment question.
If your company could eliminate 70% of today's customer-initiated service interactions over the next decade, would the technology you're buying today become more valuable or less valuable? And if your customers' personal AI agents eventually understand their priorities better than your own customer models do, how would that change your investments in CRM, personalization, analytics, and customer experience management?
I think enterprises should be asking both questions now, particularly before committing tens of millions of dollars to platforms that may shape their technology and operating models for years. Voice AI may turn out to be one of the most important advances in contact center technology, while also helping bring about a future in which much of today's contact center becomes unnecessary.
And the bigger competition may ultimately be over who has the most useful understanding of the customer, who is trusted to act on that understanding, and who gets to influence what happens next. That's a much larger opportunity than automating a phone call.
Explore Be Customer Led's independent CX technology research
Native AI Conversational Platforms Built for Resolution is a 49-page evaluation of 16 vendors covering the capabilities, operating requirements, and practical limitations of AI platforms built to resolve customer service work.
The CX Technology Market provides independent analysis of more than 70 vendors across 12 buying categories, including market overlap, vendor selection, and practical implications for enterprise technology investments.
CX Technology Investment & Architecture Index 2027 includes a research report and buyer workbook designed to help enterprises decide which CX technology investments to fund, maintain, consolidate, reduce, or exit.
Bill Staikos is Managing Partner of Be Customer Led, an independent advisory and research business focused on customer experience, analytics, AI, and technology investment.
