Your Market Is a Job, Not a Product Category: How AI Disrupted Chegg

In early 2021, Chegg's market cap was around $14 billion. Today it is $84 million.

For years, Chegg was the default way students got help with homework. It sold step-by-step textbook solutions and expert Q&A on subscription. Then ChatGPT arrived and answered any homework question in seconds, at no cost. By late 2025, Chegg's revenue, traffic and workforce were all shrinking. The company brought its founder back as CEO and still could not find a buyer. In its 2025 Annual Report, Chegg told investors that students increasingly treat ChatGPT as a substitute for its product.

Chegg defined its customer and its market in a way that AI could make obsolete, and AI did. Jobs-to-be-Done offers a different definition, one that would have let Chegg use AI against its competitors. The approach has four parts:

1. Define your customer and your market correctly.
2. Find the unmet needs inside the customer's job.
3. Benchmark every competitor on the speed and accuracy with which it satisfies each need.
4. Aim new technology at the needs no one satisfies yet.

Start with two questions

Who is your customer, and what is your market? A company that doesn't get this right risks getting displaced by a new technology instead of benefiting from it.

Your customer: three roles, one that matters most

Most teams define the customer as "the user" or "the buyer." Every market has three types of customers, and teams that confuse them make poor product decisions:

• The job beneficiary gains when the job gets done. For Chegg, that is the student, who benefits from mastering the material.
• The job executor helps get the job done on the beneficiary's behalf. For Chegg, that is the network of subject-matter experts who wrote the step-by-step solutions.
• The purchase decision maker controls the money. In a consumer market like this one, that is often the student or a parent.

The job beneficiary matters most, because they have the most to gain. Markets also follow a pattern that product leaders should plan for:

Markets evolve to remove the job executor so that the beneficiary can get the job done on their own. AI removes executors across a wide range of jobs.

Chegg's business sold students access to answers produced by human experts. Once a student could get those answers directly from an AI, for any problem, in seconds, the human executor was no longer necessary. Chegg had built its business around the executor, and its revenue declined as students stopped needing one.

A company that serves the job beneficiary and their unmet needs serves the most durable customer in the market. A company that serves the executor or the transaction is exposed to AI.

Define your market by the job

Many companies define their market by their product. Kodak struggled to move beyond the film market, BlackBerry the keyboard market and Britannica the encyclopedia market. Each was a product-based market tied to a specific technology.

Chegg defined its market as human-expert homework help. That definition includes the solution and the executor, so it expires when they do.

The student and their job stayed the same. Human experts and AI tutors are both tools for getting that job done. A well-defined market has three parts:

1. A job beneficiary (a student),
2. pursuing a job-to-be-done (master the course material),
3. stated as a high-level goal: an action and an object, with no technology in it.

Now, more than ever, technology changes faster than anything else in your business, and the beneficiary's job changes slowest. A roadmap built on the job stays valid as technology changes, and each new technology becomes another way to get the job done better.

Find the unmet needs inside the job

Knowing the job does not tell you what to build or what to say. For that you need the unmet needs inside the job, which requires a precise definition of a customer need.

In Jobs-to-be-Done, a customer need is an action plus a variable in the job, independent of any product. The action is a verb the customer performs, such as determine, identify, diagnose, verify or interpret. The variable is the thing in the job they act on. Because every customer wants the job done quickly and correctly, we measure two things on every need: the speed of the action and the accuracy of the variable.

You find needs by mapping the steps of the job. Every job has five kinds of steps: Plan, Execute, Assess, Revise and Conclude. For a student trying to master course material, a few of the needs look like this:

A job like this has 50 to 100 such needs. Finding all of them matters, because the needs you do not serve today are openings a competitor can use to take your customers.

Next, find which needs are unmet: those where the beneficiary struggles most because current solutions are slow, difficult to use or error-prone.

A Chegg expert could help with many of the student's needs, but slowly. Chegg's Expert Q&A was a submit-and-wait service. A student posted a question or a photo of one, and a human replied in 30 minutes to a couple of hours, sometimes a full day. Students were capped at roughly 20 questions a month, and each follow-up meant another wait. ChatGPT satisfied the same needs in seconds, conversationally, with effectively unlimited follow-up questions.

Price and features mattered less than speed. ChatGPT reduced a wait of 30 minutes or more to a few seconds, and students switched. Satisfying a need faster, with at least equal accuracy, creates customer value.

Base messaging on the customer's struggle

Unmet needs also shape messaging. Many teams have responded to AI competitors by adding "AI-powered" to their homepage. That phrase tells customers nothing about why the product is valuable.

Effective messaging has three parts:

1. The job the product gets done: master the course material.
2. The functional and emotional struggle of getting it done: 11 p.m., stuck on a problem, no idea what you are doing wrong, an exam in the morning.
3. A demonstration that the product ends that struggle, quickly and accurately.

When your roadmap is a list of unmet needs, your messaging can describe the customer's struggle instead of your product's specifications. An AI competitor typically satisfies a handful of needs. Mapping the rest of the job shows the high-struggle needs no one satisfies yet, where you can deliver more value than the competition.

Benchmark competitors on speed and accuracy

Once you know which unmet needs to target, you have to satisfy them better than competitors do. Most teams compare feature lists and try to make theirs longer. This rarely works, because customers want the job done faster and more accurately, not more features. A team that copies a competitor's feature ships it after the competitor has moved on to the next one.

Jobs-to-be-Done competitive analysis works differently in two ways:

• A competitor is any solution that helps get a job step done, including companies outside your category.
• A competitor is weak where it satisfies a need slowly or inaccurately. Because a need is an action plus a variable, we measure each competitor on the speed of the action and the accuracy of the variable. That measurement sets the benchmark your solution must clear before customers will switch.

The need where they compete directly

Start with the need where Chegg and the AI tutor meet directly: determine the steps to solve a problem.

Chegg had two paths. If a student's exact problem was among its roughly 120 million archived answers, the library returned a vetted step-by-step solution in seconds. For anything else, Expert Q&A routed the problem to a human, who answered in 30 minutes to a few hours, with a cap of about 20 questions a month. Accuracy was high either way.

The AI tutor lets a student paste any problem and get steps in seconds, unbounded by a library. Accuracy is high on routine problems, but it can produce confident, wrong steps on hard or novel ones.

The AI tutor changed the benchmark on this need from searching a fixed library or waiting for a human to getting steps for any problem in seconds. Little room remains to improve on that, so a competitor cannot win this need by being slightly faster or more accurate than ChatGPT. Chegg tried anyway. It rebuilt its core product to answer new questions with AI, matching the competitor on the need it had already won. Analysts have said this slows the decline without reversing it.

The whole job

A more useful analysis benchmarks every need in the job, including the ones the AI competitor does not advertise.

On diagnosing errors, both solutions can do the work, but the AI tutor is much faster, and speed is what drew students to it. A new solution would have to beat that benchmark.

On verifying mastery, no competitor performs well. A need that no one satisfies is where a new solution can win. Comparing solutions on the single need an AI competitor advertises makes it look hard to beat; comparing them across the whole job shows where it is weak.

Aim AI at the needs no one serves

A common response to an AI competitor is to add its capability to the existing product and match it on the one need it is winning. The result usually ships late and performs worse on that need. The roadmap goes to a benchmark the company cannot beat, and the rest of the job goes unaddressed.

Chegg did this when it rebuilt its core product to answer questions automatically with AI. The product changed from human experts to AI, but the market definition did not, and the decline continued.

Most companies will use AI; the decision is which needs to aim it at. A company can aim AI at the need a competitor already serves or at the high-struggle needs no one serves yet. Here is an approach to reach product-market fit by serving high-struggle needs:

1. Target getting the whole job done, matching competitors on the needs they satisfy and outperforming them on the needs they do not.
2. Create an agentic platform that gets the whole job done faster and more accurately than the competition.
3. Refine the platform by analyzing customer usage and feedback against the needs in the job continuously satisfying more needs faster and more accurately than the competition.

Of course, there are devils in the details but keeping your agentic roadmap focused on the job-to-be-done positions you well against an AI competitor. AI is becoming inexpensive and widely available, so access to it is not an advantage; knowing which needs to aim it at is. A new competitor usually aims it at one obvious need.

The job map stays stable as technology changes. It shows where the beneficiary struggles, and it gives the team a basis for aiming new technology at those needs rather than at a competitor's feature.

Summary

The approach has four steps:

1. Define your market as the job beneficiary and their job. AI then becomes a tool for getting that job done.
2. Find the unmet needs, each an action plus a variable, where the beneficiary struggles. Use them to set the roadmap and the messaging.
3. Benchmark every competitor on the speed and accuracy with which it satisfies each need.
4. Aim new technology at the needs no one satisfies, and get the whole job done.

Apply this to your market with thrvMCP

thrvMCP identifies your customer's job, job steps, customer needs and competitor weaknesses in minutes. It connects that data to the AI assistant you already use, so you can apply the Jobs-to-be-Done lens to product, marketing and sales work.

With the job data connected, you can ask your AI to prioritize a roadmap around unmet needs, write messaging that describes the customer's struggle, or prepare for a sales call by reviewing where a competitor is weak.

With thrvDisrupt, you can use thrvAI to build and iterate on your new agentic platform, using agentic business models priced against the value you deliver your customers.