I came to work out.
My phone didn’t.

At a gym, the immediate task is to confirm membership and enter. The larger purpose might be exercise, recovery, company or an hour away from demands. A check-in process should support that purpose.

If the design brief starts with “increase app adoption,” a mandatory app can look successful. If it starts with “help members begin their visit smoothly,” the same decision deserves a different examination. How many people needed help? How long did entry take? Could someone arrive without a phone and still use the membership they had paid for?

This applies to a restaurant guest trying to order lunch, a museum visitor trying to enter with children, a hotel guest arriving after a long journey and a customer trying to resolve an incorrect bill. None of them necessarily came to participate in your software rollout.

Write the customer’s intended outcome in plain language before choosing the channel. Then identify the legitimate requirements: eligibility, identity, payment, safety, accurate records or a combination. Technology should meet those requirements while making the task easier to complete.

There are several reasons a person may not use the preferred digital route. They may lack a suitable device, have a dead battery, encounter an inaccessible interface, struggle with the process, distrust the data request or simply want to disconnect.

These are different situations. A tutorial helps someone who wants to learn. It does little for someone who deliberately left their phone at home.

The UK Government Digital Service’s guidance on researching people who do not use digital services recommends investigating why people avoid a digital service. That research principle is useful well beyond government: ask what is stopping this person, instead of assigning everyone the same explanation.

Do not turn age into a shortcut for that research. A younger customer may avoid an app; an older customer may prefer it. The same person may choose different channels depending on the task, the circumstances and the stakes.

At the YMCA, my preference was not a request for technology training. I wanted some me time. A business that understands that distinction has a better chance of designing an experience people want to repeat.

Ticketing systems can improve service. They preserve a request, route it to the right team, document actions and keep an issue from disappearing when an employee leaves for the day.

They fail customers when the business treats creating the record as completing the work. “Your request has been received” answers a narrow question. It does not explain who will act, what happens next or how an urgent exception gets attention.

Consider a fictional customer whose paid admission will not scan. An online form that produces a ticket for review next week may work for an ordinary account question. It does not solve being unable to enter today. The business needs an approved on-site exception path, with the same appropriate verification and a clear owner.

For less urgent cases, name the next step, offer a realistic response expectation and keep the customer informed when it changes. Give staff authority to resolve defined issues and a route to escalate those they cannot resolve.

Your CRM decision should follow those operating needs. Buying a more elaborate system will not establish responsibility that the organization has never assigned.

AI can help answer routine questions, collect relevant facts and prepare a useful summary for a person. It can also become another locked door if the only response to a failed answer is an invitation to ask the same question differently.

Design the handoff before launch. Decide which issues require a person, what signals should trigger escalation and how customers request it. Make the next route visible. Do not promise immediate human help if no one is available to provide it.

A good transfer includes the issue, relevant steps already attempted and information the customer has permission to share. It should avoid requiring the customer to repeat the entire story. Staff still need to verify important facts; a generated summary is a working aid, not unquestionable evidence.

Be explicit about sensitive matters. A chatbot should not improvise identity checks, safety decisions, refund authority or clinical guidance beyond its approved role. The alternative channel must preserve the controls appropriate to the task.

The goal is not to make the bot sound endlessly sympathetic. The goal is to help the customer finish the task or reach someone who can act.

Use this tool in a service-design meeting with the people who actually run the desk, answer requests and administer the system. Choose one scenario, walk through it and record the point where the customer can no longer proceed.

Do the walk-through without a staff member quietly fixing the problem offstage. If a supervisor must use a special workaround every time, that is part of the process and its cost. If the workaround bypasses a necessary control, it is not a finished solution.

Test the alternative under ordinary pressure as well as in an empty room. A usable route needs instructions, staffing and enough capacity for the times customers will actually need it. The Government Digital Service’s assisted digital support guidance includes staff-assisted routes and emphasizes researching and measuring support. A private business must design its own appropriate model, but the emphasis on a functioning service is valuable.

A business can reduce one visible expense while creating less visible costs elsewhere. An employee spends less time on routine entry but more time resetting accounts. A shorter call queue coincides with customers giving up. Fewer support tickets arrive because the form is too difficult to complete.

Treat these as hypotheses to investigate, not automatic explanations. Compare the whole journey using consistent definitions.

For a fictional pilot, suppose an alternative check-in route handles 200 visits a month and adds 45 seconds of staff time per visit. That is 150 minutes, or 2.5 hours, before other costs. At an assumed loaded labor cost of $30 an hour, the direct labor estimate is $75 a month. Credential costs, training, queue effects and system support could change the result.

This example does not prove an alternative is inexpensive everywhere. It shows how to replace “we cannot afford options” with an actual calculation. Compare that cost with observed demand, service failures and the value of the customer relationship. Do not invent a retention benefit to make the spreadsheet work.

Track completion, time to resolution, repeat contact, abandoned attempts and use of the alternative. Ask customers who failed as well as those who finished. App downloads alone cannot tell you whether the service improved.

An option hidden behind an argument is not a well-designed option. Tell people what they can do before they get stuck, and give staff language they can use without apologizing for the customer’s preference.

A proposed sign for a facility with these verified capabilities might say: “Use your app or membership card. Need help checking in? Please speak with the desk.” Publish that only if all three routes work.

For support, say what the channels actually provide: immediate self-service for certain changes, staffed help during stated hours, and an appropriate route for time-sensitive problems. Avoid implying round-the-clock human service when the only overnight response is an acknowledgment.

A first-visit museum experience illustrates why this belongs in marketing too. The arrival instructions, admission process and help desk form one experience. Promotional language cannot repair a confusing entrance after the visitor arrives.

You do not have to preserve every old channel forever. You do need to understand who uses it, why, what replaces it and whether that replacement works before retiring it.