Ted TschoppSome of the clearest lessons about work do not arrive at work. They show up at the kitchen counter on a Tuesday night, when our 10-year-old twin girls are both talking at once, our 12-year-old son is looking for a signed form, and somewhere under the pile is the one school email I should have read earlier.
One of the girls is trying to figure out whether “due tomorrow” means before school or before midnight. Her sister suddenly remembers she needs poster board for a project that has apparently existed for two weeks but only became urgent right now. Our son needs clothes washed, a ride figured out, and that form signed before morning.
In a moment like that, I do not need “technology.” I need progress.
First, I need clarity. What matters most? What can wait until tomorrow? What is the next right step? Then, once the fog clears, I need follow-through. Put it on the calendar. Set the reminder. Order the poster board. Make sure the things we decided actually happen.
That same pattern shows up at work more often than we admit. And it is one of the simplest ways to think about AI.
Most conversations about AI begin with the tool. Should this be a chatbot? Should this be an agent? Should this be automation?
But most employees are not waking up thinking, I hope I get to use AI today.
They are thinking, I need to find the answer, finish the draft, get the approval, close the ticket, and move on with my day.
That is where Jobs to Be Done helps.
Start with the job, not the tool
“Jobs to Be Done” sounds like a phrase that wandered out of a strategy meeting and into regular life. But the idea is simple.
A job is the progress a person is trying to make in a specific situation.
Not the tool. Not the feature. Not the demo.
The progress.
So the job is not “use an AI chatbot.” The job is more like:
- help me understand this policy before I make a mistake
- help me compare my options without reading twelve documents
- help me draft the first version so I am not starting from zero
- help me move this request through the system without babysitting it
That is what needs rewiring in our heads. Once you see it, a lot of AI discussion gets less foggy.
A simple way to write the job is this:
When I am in this situation, I want this help, so I can make this progress.
For example:
When I am trying to get access to a system before a customer meeting, I want clear guidance and the right actions taken, so I can do my job without losing half a day to tickets and approvals.
That is a real job. It has a situation, a need, and an outcome.
The question that makes chatbot versus agent clearer
Once you can name the job, the next question is surprisingly simple:
Is the main problem uncertainty, or is it effort?
Sometimes the biggest problem is uncertainty. The employee is not sure what something means, which policy applies, what the next step is, or how to begin.
Sometimes the biggest problem is effort. The path is mostly known, but the work is repetitive, spread across systems, full of handoffs, and easy to drop.
That is where the difference between chatbots and agents becomes useful.
- AI chatbots are hired to reduce uncertainty.
- AI agents are hired to reduce effort and ensure follow-through.
A chatbot mostly says: help me think. An agent mostly says: help me do.
That is not the whole story, but it is a very good place to begin.
What a chatbot is
When I say chatbot here, I mean a conversational AI tool that helps someone understand, decide, summarize, draft, or learn.
A chatbot is a good fit when the employee is really asking questions like these:
- What does this mean?
- What policy applies here?
- Which option makes the most sense?
- Can you summarize this for me?
- Can you help me draft a first pass?
- Can you coach me through the next step?
In a large enterprise, that often means things like:
- finding answers buried in internal knowledge
- explaining policies and procedures in plain language
- helping draft emails, summaries, documents, or justifications
- supporting onboarding and training
- helping someone get unstuck when they are unsure what to do next
In simple terms, chatbots help people decide and express.
What an agent is
When I say agent here, I mean software that can take action inside a workflow on someone’s behalf, usually within rules, approvals, and systems the company already uses.
That part matters.
An agent is not magic. In most enterprise settings, it works best when the path is fairly clear, the actions are bounded, and a human can step in when something unusual happens.
An agent is a good fit when the employee is really asking:
- Can you file this?
- Can you route this for approval?
- Can you watch this and tell me if something changes?
- Can you update the right systems?
- Can you follow up until this is done?
- Can you handle the normal path and bring me in only for exceptions?
In a large enterprise, that often means things like:
- opening and routing requests
- tracking approvals
- monitoring deadlines or service levels
- updating records across systems
- coordinating follow-up steps
- keeping routine work moving without constant checking
In simple terms, agents help people complete and coordinate.
A practical way to decide: chatbot, agent, or both
Here is a simple process you can use.
1. Name the situation
Start with a real moment in someone’s work.
Are they onboarding a new hire? Trying to understand a policy? Buying software? Chasing an approval? Handling a customer issue? Troubleshooting a system problem?
Be specific. “Improve productivity with AI” is too vague. “Help managers get software requests through procurement faster” is much better.
2. Write the job in plain language
Use the basic template:
When I am… I want… so I can…
This forces the conversation out of tool language and back into human language.
3. Find the real friction
Ask what is making the work hard.
Is the person confused, overwhelmed, or unsure? That points toward a chatbot.
Is the work repetitive, spread across tools, dependent on reminders, or easy to drop? That points toward an agent.
Is it both? Then you may need both.
4. Choose the starting pattern
A simple rule of thumb:
- If the pain is mostly uncertainty, start with a chatbot.
- If the pain is mostly manual effort and follow-through, start with an agent.
- If the work begins in uncertainty and ends in execution, you probably need both.
5. Add enterprise reality
In a large company, helpful is not enough. The solution also needs to be safe, consistent, and visible enough that the organization can trust it.
That means asking questions like:
- Which systems are involved?
- What approvals matter?
- Where does a human need to review the work?
- What should be logged or tracked?
- What should happen when something unusual occurs?
At enterprise scale, speed alone is not enough. Fast and messy is still messy.
6. Decide how success will be measured
For chatbots, success might look like faster answers, better drafts, shorter time to understanding, or fewer basic support questions.
For agents, success might look like faster completion, fewer dropped handoffs, fewer manual touches, lower rework, or better compliance.
Those are different kinds of value. They should be measured differently.
Walk through one example
Let’s take a familiar situation.
An employee needs access to a system before an important customer meeting.
Here is the job:
When I am preparing for a customer meeting and realize I do not have access to the system I need, I want clear guidance and the request moved forward quickly, so I can do my job without wasting time on confusion and follow-up.
Now look at the friction.
Part of the problem is uncertainty:
- Which access level do I actually need?
- What is the right process?
- Is approval required?
- What information do I need to provide?
That is a chatbot problem.
A chatbot can explain the policy, tell the employee which request path fits, summarize the requirements, and even help draft the business justification.
But part of the problem is also effort:
- open the request
- route it to the right approver
- track the status
- remind someone if it stalls
- notify the employee when it is done
That is an agent problem.
An agent can handle the standard workflow, keep an eye on the handoffs, and pull a person in only when something unusual happens.
This is why the strongest solutions are often hybrids.
The chatbot helps the employee understand what to do. The agent helps the work keep moving. First clarity. Then execution.
A few familiar enterprise examples
The same pattern shows up all over a large company.
HR and onboarding
A chatbot can explain onboarding steps, answer benefits questions, summarize policy, and draft a welcome note.
An agent can create tasks, request access, schedule reminders, track completion, and escalate missing approvals.
IT support
A chatbot can help troubleshoot a problem, explain likely causes, and point the employee to the right process.
An agent can open the ticket, route it correctly, check status, update the case, and notify the employee when something changes.
Procurement or finance
A chatbot can explain spending rules, compare options, and draft the business justification.
An agent can create the request, collect the required fields, route approvals, track status, and update the procurement system.
Customer service or operations
A chatbot can summarize a case, suggest a response, and highlight the next best step.
An agent can log updates, send follow-ups, monitor deadlines, and escalate when service commitments are at risk.
In each case, the pattern is the same.
The chatbot helps the person make sense of the work. The agent helps the work keep moving.
Why the best answer is often both
This is the part many teams discover a little later.
The best enterprise solutions are often not chatbot-only or agent-only. They are a handoff.
The chatbot handles the front end of the experience: clarification, discovery, drafting, trust-building, and approval.
The agent handles the back end: execution, monitoring, coordination, and follow-through.
One helps the employee know what to do. The other helps the organization actually get it done.
That is usually where the real value shows up.
The simplest way to remember it
A useful shorthand is this:
- Chatbots reduce uncertainty.
- Agents reduce effort and ensure follow-through.
Or even shorter:
- Chatbots help me think.
- Agents help me do.
And one more line is worth keeping close:
- Start with the job, not the tool.
Because once you name the job clearly, the right pattern becomes easier to see.
I keep coming back to that kitchen-counter moment with our twin girls, our son, the school emails, the permission slips, and the low-grade chaos that comes with trying to get five people to tomorrow in one piece.
One daughter needs help understanding what the assignment actually means. Another needs the poster board to somehow show up before morning. Our son needs the form signed and the ride nailed down. First I need clarity. Then I need follow-through.
That is not a bad way to think about AI at work either.
Some problems begin with confusion. Some begin with effort. Many have both. Chatbots help people make sense of the moment. Agents help keep the work moving once the path is known. And the best designs often let one hand off to the other.
So before asking, “Where can we use AI?” it is worth asking the better question:
What job are we trying to help someone get done?
That question usually clears more fog than people expect.
