Table of Contents
- Key takeaways
- What is AI-enabled customer experience?
- The three layers of AI in customer experience
- How AI works at each stage of the customer journey
- 1. Discover: answering questions before a purchase
- 2. Buy: purchase questions and lead capture
- 3. Onboard and use: setup and how-to help
- 4. Get help: resolving issues and handing off
- 5. Stay and renew: finding problems before customers leave
- The four-stage AI CX maturity model
- What does the evidence say about AI-enabled CX?
- Worked example: what a 14% productivity gain means
- Where AI-enabled CX goes wrong
- How to choose where to start with AI in CX
- Where SiteGPT fits in AI-enabled customer experience
- Frequently asked questions
- Basics
- What is AI-enabled customer experience?
- Is AI-enabled CX the same as AI customer service?
- What are examples of AI-enabled customer experience?
- Journey and maturity
- Which stage of the customer journey should use AI first?
- What does a mature AI CX program look like?
- How long does it take to move up a maturity stage?
- Risks and trust
- Do customers want AI in customer experience?
- Does a chatbot have to say it is AI?
- Should AI replace human agents?
- Cost and setup
- What does AI-enabled customer experience cost to start?
- What do you need before adding AI to CX?
- Sources
Created by
Do not index
Created time
Oct 17, 2023 11:28 PM
AI-enabled customer experience means using AI at each stage of the customer journey to answer, guide, and route customers, with people handling what the AI cannot resolve.
TL;DR
- AI-enabled customer experience (CX) covers five journey stages: discover, buy, onboard and use, get help, and stay and renew.
- AI works in three layers: customer-facing (chatbots, email, messaging), agent-facing (drafted replies, summaries), and insight (topic and sentiment tagging).
- Most teams move through four maturity stages: manual, answering, assisted and measured, and proactive. Skipping a stage usually fails.
- The evidence is real but narrow. AI assistance raised issues resolved per hour by 14% in an NBER study of 5,179 agents, while only 27% of customers would retry a chatbot after a bad experience (Gartner, 2026).
- Start where the same questions repeat and the answers already exist in your content. Keep a visible route to a person at every stage.
Key takeaways
Journey stage | What AI does | Keep human for | Metric to watch |
Discover | Answers product, pricing, and fit questions at any hour | Custom quotes and complex fit questions | Questions answered without a person |
Buy | Handles purchase questions and captures leads | Negotiation and large deals | Leads captured per 100 conversations |
Onboard and use | Walks customers through setup and how-to questions | Account-specific setup problems | Time to first value |
Get help | Resolves routine issues and hands off the rest with context | Exceptions, refunds, upset customers | Resolution time and repeat contacts |
Stay and renew | Tags topics and sentiment, flags recurring problems | Renewal and save conversations | Churn and recurring complaint topics |
What is AI-enabled customer experience?
AI-enabled customer experience is a customer experience in which AI handles or supports parts of every interaction a customer has with a company. That includes questions before a purchase, the purchase itself, setup, service requests, and renewal.
Customer service is one part of this. AI customer service focuses on the "get help" stage. AI-enabled CX asks where AI helps across the whole journey, and where it should stay out.
McKinsey describes AI-enabled service as a way to deliver personalized, proactive experiences, and it stresses that "AI-powered does not mean automation-only." People stay in the model for complex needs and for moments that build the relationship.
The three layers of AI in customer experience
AI shows up in three layers. Most companies start with one and add the others later.
Layer | Who uses it | Examples | What it changes |
Customer-facing | Customers | Website chatbot, AI email replies, messaging app bots | Customers get answers without waiting for a person |
Agent-facing | Service agents | Drafted replies, chat summaries, suggested next steps | Agents resolve more cases per hour |
Insight | Managers and product teams | Topic tagging, sentiment analysis, unanswered-question reports | Teams see why customers get in touch and fix the cause |
For a full list of the tasks AI can take on inside a service team, such as triage, translation, and QA, see the guide to customer service AI use cases.
How AI works at each stage of the customer journey
AI works differently at each stage of the customer journey because customers need different things at each one. Each stage below follows the same template: what the customer needs, what AI does, an example, what to keep human, and the metric to watch.
1. Discover: answering questions before a purchase
At the discover stage, customers compare options and look for answers about features, pricing, and fit. AI answers those questions on the spot, at any hour, from the company's own website and documents.
- Customer need: a fast, specific answer without booking a call.
- What AI does: a chatbot trained on product pages, pricing, and FAQs answers in the customer's own words.
- Example: a visitor asks at 11 p.m. whether a plan includes a Slack integration and gets the answer from the pricing page.
- Keep human: custom quotes, security reviews, and unusual fit questions.
- Metric: share of pre-sale questions answered without a person.
2. Buy: purchase questions and lead capture
At the buy stage, customers have a final question that stands between them and the purchase. AI answers it or captures their details so a person can follow up.
- Customer need: confidence to buy now, or a quick route to sales.
- What AI does: answers delivery, billing, and plan questions, and asks for contact details when the visitor wants a callback.
- Example: a shopper asks about the returns window, gets the policy, and completes checkout. The guide to chatbots for ecommerce covers store use cases.
- Keep human: negotiation, large orders, and contract terms.
- Metric: leads captured per 100 conversations.
3. Onboard and use: setup and how-to help
At the onboard and use stage, customers want to get value from what they bought. AI answers setup and how-to questions from help articles, videos, and docs.
- Customer need: step-by-step help at the moment of confusion.
- What AI does: finds the right help article and explains it in plain language, in the customer's language.
- Example: a new user asks how to connect an account and gets the three steps from the help center.
- Keep human: problems tied to a specific account or a bug.
- Metric: time to first value, and how-to questions answered without a ticket.
4. Get help: resolving issues and handing off
At the get help stage, customers have a problem. AI resolves routine issues in one exchange and hands the rest to a person with the conversation attached.
- Customer need: a fix, or fast access to someone who can fix it.
- What AI does: answers policy and troubleshooting questions, and escalates when the customer asks or the answer is not in its content.
- Example: a customer asks why a payment failed. The bot explains the common causes, then offers a person when the customer says none apply.
- Keep human: exceptions, refunds, complaints, and upset customers.
- Metric: resolution time, repeat contacts, and escalation rate.
The six CX metrics this stage moves are covered in detail in how AI improves customer experience.
5. Stay and renew: finding problems before customers leave
At the stay and renew stage, the goal is to keep customers. AI helps by reading every conversation and showing which problems keep coming up.
- Customer need: problems fixed for good, and no surprises at renewal.
- What AI does: tags conversations by topic and sentiment, and lists questions the bot could not answer.
- Example: tagging shows that 1 in 10 conversations this month mention the same export bug. The product team fixes it before renewals.
- Keep human: renewal calls, save offers, and account reviews.
- Metric: recurring complaint topics and churn rate.
The four-stage AI CX maturity model
AI CX maturity is how far a company has moved from people answering everything to AI handling, supporting, and anticipating customer needs. The four-stage model below is a simplified way to place a team and pick the next step.
Stage | What AI does | What people still do | Ready for the next stage when |
1. Manual | Nothing. People answer every question. | All conversations | The same questions repeat every week and help content exists |
2. Answering | A chatbot grounded in your content answers routine questions on one channel | Complex issues, handoffs, content upkeep | The bot resolves a targeted set of questions reliably |
3. Assisted and measured | Drafts agent replies, summarizes chats, tags topics and sentiment, covers more channels | Judgment calls, exceptions, relationship work | Conversation data is connected across channels |
4. Proactive | Reaches out before the customer asks, using account and behavior data | Oversight, escalations, sensitive cases | Ongoing: review outcomes every week |
McKinsey uses a five-level scale for AI-driven customer service. Level one is manual, paper-based, and high-touch. At level five, companies handle more than 95% of service interactions through AI and digital channels. McKinsey says the most mature companies tend to be digital natives in ecommerce, ride hailing, and streaming, and that some banks, telecoms, and insurers have reached levels three and four.
Most small and mid-sized teams sit at stage 1 or 2 of the four-stage model. Stage 2 is the cheapest step up because it reuses content the team already has.
What does the evidence say about AI-enabled CX?
The evidence says AI makes service teams faster and can cut contact volume, but customers drop chatbots quickly after one bad experience. Three sources carry most of the weight.
Source | What was measured | Result |
NBER working paper 31161 (Brynjolfsson, Li, Raymond), published in the Quarterly Journal of Economics in 2025 | 5,179 customer service agents given a generative AI assistant | 14% more issues resolved per hour on average, 34% for novice and low-skilled agents, improved customer sentiment |
McKinsey, March 2023 (case reported by McKinsey) | A fast-growing Asian bank's 12-month AI-enabled service program | Self-service use doubled to tripled, service interactions fell 40-50%, cost-to-serve fell more than 20% |
Gartner survey of 3,566 B2B and B2C customers, February and March 2026 | Customer attitudes to service chatbots | 49% would use a chatbot, 7% used one in their latest interaction, 27% would retry after a bad experience, 87% call human access essential |
Worked example: what a 14% productivity gain means
A team handles 1,500 chats a month at 8 minutes each. That is 12,000 minutes, or 200 agent hours a month.
- At the NBER average (14% more issues per hour): the same volume takes 200 / 1.14 = 175.4 hours. That saves about 25 hours a month, a 12% cut.
- For a team of newer agents (34% gain): the same volume takes 200 / 1.34 = 149.3 hours. That saves about 51 hours a month, a 25% cut.
The gain comes from agent-facing AI. A customer-facing chatbot adds a separate saving on the conversations it resolves without a person.
Where AI-enabled CX goes wrong
AI-enabled CX goes wrong when AI blocks access to people, answers from the wrong source, or launches everywhere at once. These five failure modes cause most of the damage.
- AI as a wall. Gartner found 87% of customers say access to a human agent is essential when companies use GenAI. Hide the person and trust drops.
- One bad answer. Only 27% of customers would try a chatbot again after a negative experience. A bot that guesses loses most of its users. The guide to stopping chatbot hallucinations covers the guardrails.
- Too much scope too soon. Gartner's advice is to "prioritize reliability over reach" and expand only after performance is reliable.
- Hidden AI. Article 50 of the EU AI Act requires that people be told they are interacting with an AI system, unless it is obvious. The Act lists the article as applying from 2 August 2026.
- Stale content. A grounded bot repeats what its sources say. Outdated help pages produce confident, wrong answers.
The chatbot best practices guide turns these into 15 practices, each with a check.
How to choose where to start with AI in CX
Start with the stage of the journey where the same questions repeat and the answers already exist. That keeps the first project small and measurable.
If you need... | Pick... | Why |
Answers to repeat pre-sale or how-to questions at any hour | A customer-facing chatbot trained on your website and help content | It reuses content you already have and covers the discover, buy, and onboard stages |
Faster agents on complex cases | Agent assist inside your helpdesk | The NBER study measured 14% more issues resolved per hour |
To know why customers get in touch | Topic and sentiment tagging on past conversations | It shows which questions to automate first |
Answers in email or messaging apps, not only on the website | A chatbot that connects to those channels | Customers keep the channel they already use |
To reach customers before they ask | A proactive messaging tool connected to account data | Proactive outreach needs stages 2 and 3 working first |
Customers to reach a person at all | Live chat staffing first, AI second | AI on top of an unstaffed channel raises effort. See chatbots vs live chat |
For choosing a platform once the starting point is clear, the guide to conversational AI for customer experience compares the options, and the longer list of conversational AI chatbots covers more tools.
Where SiteGPT fits in AI-enabled customer experience
SiteGPT is a customer-facing AI chatbot trained on your own business content. It covers stage 2 of the maturity model and parts of stage 3, across the discover, buy, onboard, and get help stages of the journey.
- Content sources: websites, sitemaps, files, raw text, Google Drive, Dropbox, OneDrive, Box, Notion, GitHub, and YouTube.
- Channels: the website widget on every plan. Growth and above add Google Chat, Messenger, Slack, Crisp, Freshdesk Messaging, and Zoho SalesIQ. Zendesk is Scale and above. WhatsApp and AI Email Support are paid add-ons.
- Buy stage: lead collection captures visitor details for follow-up.
- Get help stage: visitors can move to a live agent at the push of a button.
- Insight layer: Growth adds knowledge gaps and a weekly email digest of questions the chatbot could not answer. Scale adds AI conversation insights with topics, sentiment, and resolution outcomes.
- Languages: 95+.
- Limits: it is not an agent-assist tool inside your helpdesk. It has no Intercom or HubSpot integration today. Starter has no chat integrations and basic analytics only. The plan pages list no proactive outbound messaging, so stage 4 needs other tools.
Starter costs $59 a month billed monthly, or $39 a month billed yearly ($468 a year). Growth costs $129 a month billed monthly, or $79 a month billed yearly ($948 a year). Every self-serve plan has a 7-day free trial, and a credit card is required.
Frequently asked questions
Basics
What is AI-enabled customer experience?
AI-enabled customer experience is the use of AI at each stage of the customer journey, from pre-sale questions to renewal, to answer, guide, and route customers. People handle what the AI cannot resolve.
Is AI-enabled CX the same as AI customer service?
No. AI customer service covers the get help stage. AI-enabled CX covers the whole journey, including discovery, purchase, onboarding, and retention.
What are examples of AI-enabled customer experience?
Examples include a website chatbot that answers pricing questions at night, AI that drafts replies for agents, a bot that walks new users through setup, and topic tagging that shows which problems drive complaints.
Journey and maturity
Which stage of the customer journey should use AI first?
Start where the same questions repeat and the answers already exist in your content. For most teams that is pre-sale and how-to questions on the website.
What does a mature AI CX program look like?
A mature program uses AI in all three layers and reaches out before customers ask. McKinsey's top level describes companies that handle more than 95% of service interactions through AI and digital channels.
How long does it take to move up a maturity stage?
It depends on content and data readiness. McKinsey's bank example ran for 12 months across channels, measurement, and governance. A single grounded chatbot on one channel is a much smaller project.
Risks and trust
Do customers want AI in customer experience?
Many are open to it but quick to give up. Gartner found 49% would use a chatbot, but only 7% used one in their latest service interaction, and 27% would try again after a bad experience.
Does a chatbot have to say it is AI?
In the EU, yes, unless it is obvious. Article 50 of the EU AI Act requires that people be told they are interacting with an AI system. The AI chatbot compliance guide covers other rules.
Should AI replace human agents?
No. Gartner found 87% of customers say access to a human agent is essential when companies use GenAI. AI works best on routine questions, with a clear handoff for the rest.
Cost and setup
What does AI-enabled customer experience cost to start?
A customer-facing chatbot is the cheapest entry point. SiteGPT Starter costs $59 a month billed monthly, or $39 a month billed yearly.
What do you need before adding AI to CX?
You need current help content, a way to reach a person, and a baseline for resolution time and repeat contacts. Without those, there is no way to tell whether the AI helped.
Sources
Every external page below was opened and checked on October 23, 2026.
Written by
