
To reduce customer effort, stop the second contact: Gartner found only 14% of issues are fully resolved in self-service and 62% of channel switches are high-effort, so fix your help content, pass the full context on every handoff, and answer the next question before the customer asks it. The research behind the idea is blunt. A Harvard Business Review study of more than 75,000 people found that over-the-top service makes little difference, and that customers really want "a simple, quick solution to their problem." Below are seven fixes, in the order I would do them, and where an AI teammate like eesel takes work off both the customer and your team.
Where is your customers' effort coming from?
Pick the complaint you hear most often. Each one points to a different fix, and starting with the wrong one is how teams spend a quarter on a chatbot when the real problem was a handoff.
Pick the complaint you hear most to see which fix to start with.
What is customer effort, and why does it matter more than delight?
Customer effort is the work a customer does to get a problem solved: finding the right channel, explaining the issue, waiting, explaining it again, and coming back when the first answer didn't stick. The customer effort score (CES) measures it with a survey after the interaction.
The idea comes from Matthew Dixon, Karen Freeman and Nick Toman at CEB, in their 2010 HBR article, "Stop Trying to Delight Your Customers". Their study of more than 75,000 people found that going above and beyond makes little difference to loyalty, and that CES is a better predictor of loyalty than customer satisfaction or the Net Promoter Score, according to Harvard's summary of the article. The article opens with a question worth asking your own team: how often does anyone go back to a company because of its over-the-top service? You can probably think of a few examples, but not many.
I work on eesel's support team, and the practical upshot is simple: customers don't come back because of a heroic save, they come back when the first answer didn't work. So reducing effort is mostly about removing second contacts, and it is cheaper than delight, because the work you remove is your team's work too.
How do you measure customer effort?
Measure it two ways: ask, and count. The survey side is CES, sent after a ticket is resolved, which our customer effort score guide covers in detail. The counting side is the behaviour that effort leaves in your helpdesk:
| Signal | What it tells you | Where to find it |
|---|---|---|
| Repeat contacts on the same issue within 7 days | The first answer didn't resolve it | Helpdesk reporting, grouped by requester |
| Tickets that started in self-service | Your help center sent them to you | Help center search and article views before the ticket |
| Channel switches (chat to email, email to phone) | The first channel couldn't finish the job | Ticket channel history |
| "As I said" and "any update?" in customer messages | They repeated themselves or waited without hearing back | Search ticket text, or sentiment analysis |
| First contact resolution | The share solved without a second contact | Helpdesk reporting, or count reopened and repeat tickets |
The counted signals matter because the customers who gave up are the least likely to answer a survey. Pair CES with CSAT if you already run it; they answer different questions.
1. Resolve the next issue, not just this one
The first of the five tactics in the HBR article is to "reduce the need for repeat calls by anticipating and dealing with related downstream issues." A customer asking how to change their plan will next ask when the new price applies and what happens to their unused credit. Answer all three in the first reply.
To find the downstream questions, pull the second ticket that customers open within a week of a first one, and group by the first ticket's topic. The pairs that come up again and again become additions to your macros and help articles. This is also the most natural job for an AI agent that drafts replies: given past tickets, it learns which follow-up questions come with which first questions. Our guide to AI for first contact resolution goes deeper on that.
2. Make self-service resolve, not just deflect
Self-service is meant to be the low-effort path, and most of the time it isn't. A Gartner survey of 5,728 customers found that only 14% of customer service issues are fully resolved in self-service, and only 36% even for issues customers called "very simple". 73% of customers use self-service at some point, so a failed attempt is effort spent before the ticket even exists.
The same survey names the causes. 45% of customers who started in self-service said the company didn't understand what they were trying to do, and the most common failure, in 43% of cases, was that customers couldn't find content relevant to their issue. Gartner's recommendations follow from that: let reps write knowledge as part of resolving a ticket, let customers and reps flag articles that didn't help, and make one entry point, such as an AI chatbot, the most prominent way in.
In practice:
- Write the missing articles from your ticket history, starting with the most common questions that have no article. An AI knowledge base audit finds them faster than a manual review.
- Retire or fix outdated articles; detecting outdated help center content covers a way to catch them.
- Report resolution rate rather than deflection rate. A customer who leaves the help center and opens a ticket was deflected and still not helped.
SamCart is a good example of the entry-point idea. It runs one eesel agent on Freshdesk tickets and one in the website chat, both answering from 280 help center articles. They link only public help articles, and the chat asks for a name and email before it answers so the team can follow up.
3. Hand off with the context attached
Having to explain the problem twice is one of the complaints support teams hear most. In a Gartner survey of 1,492 B2B and B2C customers, 62% of channel transitions were high-effort, meaning switching channels didn't lead to resolution. 88% of journeys that start in self-service touch more than one channel, so this isn't an edge case.
The payoff for fixing it is unusually large. Seamless transitions save an average of 4 minutes of rep time per journey "because reps aren't asking customers to repeat information they have already provided," 93% of those customers reported high CSAT, and 74% said they would use self-service again next time. Fewer than half of customers with a high-effort transition said the same.
What to pass along on every handoff, from a bot to a person or from one team to another:
- What the customer asked, in their words, and what they already tried (the articles they read, the steps in the chat)
- What the bot or the first agent already checked, such as the order status or the account plan
- Why it is being handed over, in one line
Gartner's first recommendation is exactly this: capture the context in self-service, such as the FAQs a customer visited, and transfer it to the rep. Our guides to AI agent handoffs and chat handoffs cover the setup in common helpdesks.
4. Guide customers to one channel instead of offering five
Another HBR tactic is to "minimize the need for customers to switch service channels." A contact page that offers chat, email, phone, a form and a community forum looks generous, but it hands the routing decision to the person least able to make it. Gartner's advice in the 2023 survey is to give prescriptive channel guidance rather than a menu of choices: ask what the issue is, then send the customer to the channel that resolves that kind of issue.
Two practical rules: route billing disputes and anything with an account change straight to a person, and keep the way to a person visible. Customers are more willing to try self-service when they know it is easy to get more help, which Gartner also recommends telling them up front. Our glossary entry on omnichannel support has more on keeping one conversation across channels.
5. Ask your agents to recommend self-service, and mean it
This one surprises most support leads. A Gartner survey of 5,801 customers found that 60% of customer service agents fail to promote self-service, and when they do mention it, 25% make neutral comments and 12% make explicitly negative ones. Agent promotion is associated with a doubling of the number of customers likely to use self-service next time.
So when an agent resolves a ticket that the help center could have handled, the reply should point to the article that would have answered it, with a line on when it's the faster route. That only works if the article really is good, which is why this fix comes after fix 2. The same survey found only 35% of customers who last contacted a company by phone are willing to use an AI assistant, so make the AI option an addition to the channels customers trust, not a wall in front of them.
6. Go after the customers who are struggling
The fourth HBR tactic is to "elicit and use feedback from disgruntled or struggling customers." Low CES and CSAT responses are the obvious source: read every one, and contact the customer when the issue is still open. Make it routine, so the quiet failures get found too:
- Tickets with no assigned group, or pending with no follow-up date, are customers waiting with nobody on the other end.
- Customers who opened a second ticket on the same issue are the ones the first answer failed.
Cape Coffee Beans runs exactly this kind of sweep with eesel. Every morning eesel looks for Zendesk tickets with no group assigned, pending tickets with no follow-up date and enquiries on the wrong ticket form, and emails the founder a short list with a link to each. Over a month that came to 160 scheduled checks across Zendesk and Google Drive. A customer whose ticket gets picked up before they chase never has to spend the effort of chasing.
For a broader look at catching at-risk customers, see AI churn prevention for support.
7. Reward solving the problem, not handling time
The last HBR tactic is to "focus on problem solving, not speed," alongside arming reps "to address the emotional side of customer interactions." A team measured on average handle time learns to close tickets fast, and a fast close that doesn't resolve the issue is a repeat contact waiting to happen.
Swap the headline metric to first contact resolution and repeat contacts, keep handle time as a diagnostic, and let agents spend the extra minute on the downstream questions from fix 1. The emotional side matters most on the tickets where the answer is no: say why, say what the customer can do instead, and the customer is far less likely to come back to argue.
What are the most common mistakes when reducing customer effort?
- Hiding the way to a person. A help center with no visible contact option protects the deflection number and adds effort. Gartner's finding that 74% of customers with an easy handoff say they'll use self-service again is the case for keeping the door open.
- Adding a channel before fixing content. A chatbot that answers from the same missing articles hits the same gap, the top cause of self-service failure in Gartner's survey, only faster.
- Measuring speed alone. First response time and handle time can improve while repeat contacts climb.
- Surveying and not acting. A low CES response that nobody reads tells the customer their effort was wasted twice.
Where does AI fit in reducing customer effort?
AI is good at the parts of effort that come from waiting and repeating: answering at any hour from your help center and past tickets, carrying the whole conversation into a handoff, and noticing tickets that have gone quiet. It is not a fix for missing content or a policy customers hate; it will repeat both faster.
eesel is an AI teammate that works inside the helpdesk you already use, including Zendesk and Freshdesk. It learns from your past tickets and help center, is tested against your past tickets before it replies to anyone, and can draft for your team, reply on its own, or both. What it may do depends on who it is talking to: it can issue a refund when your support lead asks, ask a person first when a verified customer asks, and never touch billing for an anonymous visitor. Each tool can act on its own, ask first or stay off (actions and approvals). smava, for example, has eesel sort and draft replies for 27,000 Zendesk tickets a month, with every draft waiting for an agent's approval. If your team works with an AI coding agent, it can also do much of the setup through the eesel CLI, such as uploading approved help articles and macros or asking test questions before go-live, while connecting the helpdesk still needs an admin's approval.
Try eesel on the tickets your customers write twice
If the second contact is where your customers' effort goes, start there: connect your helpdesk, let eesel read your past tickets, and see how it would have answered the repeat questions before it talks to a single customer. It hands over to your team with the conversation attached, so nobody has to explain it twice. Try eesel.
