The 80% Customer Service Shift
Gartner forecasts that agentic AI could resolve 80% of common service issues by 2029. The important question is what resolution really requires.
Read article →Voice AI
Insights
Ideas for putting AI to work.
Analysis and guidance on AI agents, customer operations, automation and the role of human oversight.
Lead Management
130,000 Leads Nobody Was Touching
Salesforce used AI agents to contact leads its sales organization was not working, creating 3,200 opportunities in four months.
Read article →Human Oversight
The AI Paradox
Service leaders expect more automation while retaining people. The future may depend on designing the handoff between them.
Read article →Automation
88% Use AI. Only 7% Have Actually Scaled It.
McKinsey’s 2025 survey shows the gap between adopting AI tools and redesigning work around them.
Read article →AI Operations
AI Is Creating Demand for New Roles
As companies deploy agents, they also need people who can design, supervise and improve the work those agents perform.
Read article →Customer Experience
Companies Think AI Understands You. Customers Disagree.
Twilio found a wide gap between how well businesses think they understand customers and how understood customers actually feel.
Read article →Featured insight
In McKinsey’s survey, 88% use AI. Only 7% have fully scaled it.
88% use AI
AI adoption is already widespread across business functions.
Only 7% have scaled it
Most organizations remain in experimentation or partial deployment.
The bottleneck
The challenge is moving from isolated AI tools to redesigned business workflows.
The opportunity
The next advantage may come from integrating AI directly into how work gets done.
Based on McKinsey's 2025 Global Survey on AI · 1,993 participants
Read article →Voice AIThe 80% Customer Service Shift
By 2029, AI could handle 80% of common service issues. The interesting part is what happens next.
Gartner predicts agentic AI will autonomously resolve 80% of common customer-service issues by 2029, with a corresponding 30% reduction in operational costs. That doesn't mean 80% of customer service jobs disappear. It means the structure of customer service changes.
For most of the history of customer service automation, technology has helped customers find information.
Press 1 for sales. Press 2 for support. Search the help center. Ask the chatbot.
Agentic AI introduces something different: the system can potentially act on the customer's request rather than simply tell them what to do next.
That distinction is why Gartner predicts that by 2029, agentic AI will autonomously resolve 80% of common customer-service issues without human intervention, contributing to a 30% reduction in operational costs.
The interesting word is resolve. Not answer. Not respond. Resolve.
Answering a question and resolving a request are different things
Imagine a customer calls because they need to move an appointment. A traditional automated system might explain the cancellation policy or direct them to an online booking page. A conversational AI might understand the request and explain what options are available.
An AI agent can potentially go further:
Identify the customer → retrieve the appointment → check permitted availability → offer alternatives → change the booking → update the calendar → confirm the new appointment.
The conversation becomes the interface to the underlying workflow. That is a much larger change than making call-center conversations sound more natural.
Voice becomes an operating interface
For businesses, this creates an interesting possibility. The telephone has historically created work. A customer calls. An employee answers. The employee listens. Then the employee performs the actual work in another system.
They open the CRM. They check the calendar. They update a record. They send a confirmation. They create a follow-up.
If AI can safely perform some of those actions during the conversation, the economics of a call change. The call is no longer simply something the business must answer. It becomes a trigger for work that can potentially be completed immediately.
But 80% automation does not mean removing humans
In a 2025 poll of 163 customer-service leaders, 95% said they planned to retain human agents. Gartner separately predicted that by 2027, half of organizations that had expected to significantly reduce customer-service headcount because of AI would abandon those plans.
The likely transformation is not AI replaces customer service. It is: AI handles a much larger proportion of defined interactions, while humans concentrate on the interactions where judgment, authority, empathy or exception handling matters.
The strategic question is: Which customer requests can AI actually resolve from beginning to end - and which should remain human?
Sources: Gartner’s March 2025 service forecast and Gartner’s June 2025 workforce research.
Lead Management130,000 Leads Nobody Was Touching
Salesforce put AI on 130,000 untouched leads. It created 3,200 opportunities.
The surprising sales opportunity for AI may not be replacing salespeople. It may be working the leads salespeople never had enough time to contact.
Sales organizations spend enormous amounts of money generating leads. But generating a lead and having enough human capacity to pursue every lead are very different things.
Salesforce's 2026 State of Sales research reports that sales representatives spend nearly one full working day each week prospecting, yet 48% still say they lack enough bandwidth to perform adequate cold outreach.
130,000 untouched leads
Salesforce says it deployed AI agents against leads that its human sales organization was not working. Over four months, those agents contacted 130,000 previously untouched leads. The result: 3,200 opportunities created.
That works out to roughly one opportunity for every 41 leads contacted. These were leads that were effectively falling outside available human capacity.
AI changes the economics of the long tail
Human attention is expensive and finite. An AI agent can potentially handle the first layer of engagement: contact the prospect, establish whether interest still exists, identify basic requirements, determine whether the opportunity meets defined criteria and bring a salesperson into the conversation when appropriate.
AI doesn't necessarily replace the salesperson. It expands the number of opportunities the sales organization can economically examine.
Salespeople themselves are moving toward this model
Salesforce reports that 54% of sellers surveyed had already used AI agents, with nearly nine in ten planning to use them by 2027. Top-performing sellers were 1.7 times more likely than underperformers to use prospecting AI agents for outreach.
A better question than whether AI can sell as well as a person is: How many legitimate opportunities never receive sufficient attention because the sales team doesn't have enough time?
Source: Salesforce, State of Sales 2026.
Human OversightThe AI Paradox
Gartner found that 95% of service leaders planned to retain humans, while forecasting that AI could resolve 80% of common service issues by 2029.
Two of the most interesting predictions about customer service appear to contradict each other. Together, they may actually describe what the future operating model looks like.
In March 2025, Gartner surveyed 163 customer-service and support leaders. 95% said they planned to retain human agents. Gartner also predicts that by 2029 agentic AI will autonomously resolve 80% of common customer-service issues without human intervention.
Automation and replacement are not the same thing
“What time do you close?” and “You've charged me €4,000 incorrectly and I've already called three times” are both customer-service interactions, but operationally they have almost nothing in common.
The first is repetitive, low-risk and governed by known information. The second may involve frustration, financial consequences, unusual circumstances and authority to make an exception.
The future may be defined by escalation design
The important AI capability may not only be knowing what to do. It may be knowing when not to do it. A mature AI operating model needs explicit boundaries for answers, appointments, information collection, approvals, valuable prospects and unusual requests.
Humans become the exception layer
Today, humans often handle almost everything while automation assists around the edges. Tomorrow, AI may handle the predictable majority while people concentrate on exceptions.
The winning model is designed responsibility. AI handles what the organization has determined AI should handle. Humans retain authority where human judgment is required. The transition between the two is part of the system itself.
The question is whether the business can define precisely where AI's authority ends.
Sources: Gartner customer-service workforce research, Gartner’s agentic-service forecast and Microsoft’s 2025 Work Trend Index.
Automation88% Use AI. Only 7% Have Actually Scaled It.
In McKinsey’s survey, 88% use AI. Only 7% have fully scaled it.
The biggest AI problem inside companies may no longer be adoption. It may be turning isolated AI tools into actual operating processes.
McKinsey's 2025 global survey of 1,993 participants found that 88% of respondents said their organizations were using AI in at least one business function. Only 7% said AI had been fully scaled across their organizations.
Using AI is easy. Redesigning a business around what AI can do is much harder.
A chatbot is not transformation
AI can write a response to a customer requesting an appointment, while the employee still reads the response, opens the calendar, finds availability, replies, waits for confirmation, creates the booking, updates the CRM and sends another confirmation.
If the system can understand the request, check approved availability, schedule the appointment, update the relevant systems and escalate only when an exception occurs, the process itself has been redesigned.
McKinsey points directly at workflow redesign
McKinsey notes that scaling AI may require organizations to redesign workflows around AI capabilities and establish platforms capable of operating at scale.
The next question is: If this process were designed today, with AI available from the beginning, would we design it the same way?
Start with the workflow, not the technology
The best automation candidates share high frequency, clear rules, structured information, manual friction and defined exceptions. The objective is not to automate everything. It is to find processes where humans are spending time because historically there was no practical alternative.
The 7% number is the opportunity
Most organizations have crossed the experimentation threshold. Very few have completed the operational transformation. The competitive advantage may shift from who has AI to who has redesigned their business so AI can actually do useful work inside it.
Source: McKinsey, AI at work but not at scale, based on its 2025 global survey of 1,993 participants.
AI OperationsAI Is Creating Demand for New Roles
AI agents are not only changing work. They are also creating demand for new operational roles.
28% of managers surveyed by Microsoft were considering hiring AI workforce managers, while 32% planned to hire AI agent specialists within 12-18 months.
Organizations need people to manage the agents themselves. Those roles exist because AI is moving from being a tool employees use to a new operational layer businesses have to manage.
From user to manager
A person may define the objective, give the agent access to particular information, determine what actions it may take, set approval requirements, review performance, manage exceptions and improve the workflow over time.
The employee is no longer simply operating software. They are supervising digital work. Microsoft calls the emerging version of this role the agent boss.
Companies expect employees to build and manage agents
Microsoft found that leaders expected teams to redesign business processes with AI, build multi-agent systems, train agents and manage agents. Employees will need to understand responsibility, system access, success, approvals, escalation, logging and failure handling.
The org chart could eventually include digital roles
AI agents can perform defined roles, and someone has to own those roles, decide what they may do, measure their output and improve them.
The first era was: Every employee gets an AI assistant. The emerging era may be: Every team manages AI agents.
Customer ExperienceCompanies Think AI Understands You. Customers Disagree.
82% of businesses think they understand their customers. Only 45% of customers agree.
AI is giving companies more customer data and more ways to personalize interactions. Customers don't necessarily feel the difference.
Twilio's 2025 State of Customer Engagement research surveyed 7,640 consumers and 637 business leaders across 18 countries. Among business leaders, 82% said they deeply understand their customers. Only 45% of consumers agreed that brands understand them.
More intelligence doesn't automatically create more understanding
Businesses may be becoming more technically sophisticated without the customer experiencing a more coherent relationship. Possessing information and using it at the right moment are different things.
Customers experience context, not databases
Customers need businesses to know who they are when appropriate, know what already happened, avoid asking for information twice, understand why they are contacting, take the correct next action and let them reach a person when necessary.
Irrelevance has a measurable cost
Irrelevant experiences can weaken engagement. Zendesk found that 64% of consumers surveyed said they were more likely to trust AI agents displaying traits such as friendliness and empathy.
But sounding human cannot compensate for having no useful memory, giving the wrong answer or failing to complete the customer's request.
The next competitive advantage may be continuity
The best outcome may not be “I couldn't tell it was AI.” It may be: “It already knew what was happening, solved what I needed and I didn't have to repeat myself.”
Sources: Twilio, 2025 customer-engagement analysis and Zendesk, 2025 CX Trends Report.