Start With the Outcome, Not the Tool

If you have not started with AI yet, you are in a better position than you think.

Anthropic just launched reporting tools that show enterprise teams what their AI spend produced. It is a sign of where early adopters have landed: they bought fast, and now they are scrambling to prove it paid off. You can skip that step.

At TEG, the thing we care about most is value. That means starting from the outcome you need and working back to the tool.

What early adopters are running into

On October 28, Anthropic announced a suite of reporting tools built on internal analytics and employee surveys. Managers can see a team’s work over a date range, what it produced, what it cost, and where workflows broke down. The reports include recommendations, and Anthropic is adding pooled group budgets so teams can manage token spend together. By default, Anthropic says it will not train on this data.

The numbers behind the move explain the timing:

• TechCrunch, citing Jellyfish data, reports that Claude Code and tools built on Anthropic models account for nearly a third of proposed code changes, with spend up 10x since January.
• Most of the Fortune 500, per TechCrunch, are not yet seeing major gains from deploying AI.
• Jellyfish’s October 2026 report shows the median company’s cost per merged pull request fell about a third, from $1,202 to $820. AI’s direct share of that cost is about $18.

Notice which number matters most. Cost per merged PR is an outcome. Token counts are not.

Jellyfish’s chief scientist put the problem plainly: a 10x jump is growth most CFOs did not plan for, and spending more is no longer enough. You now have to optimize it.

Why a dashboard cannot replace a target

Visibility into spend is useful. Anyone paying a growing AI bill should want it. But three limits are worth naming before you lean on it.

1. Activity is not value. Tokens used, tasks completed and hours saved are inputs. A finance leader cares about outcomes: faster close, fewer errors, more deals worked, lower cost per ticket.
2. The vendor is grading its own product. TechCrunch frames these reports as a lighter version of the forward-deployed engineering help labs offer, and a Jellyfish product lead notes that switching costs rise once teams standardize workflows. Transparency and retention go hand in hand.
3. It only sees what runs through one tool. Real workflows cross your CRM, your inbox, your contracts and your people. No single vendor’s report covers that.

None of this makes the tools bad. It means the definition of value has to come from your business before any tool is switched on.

Our view: define value before you deploy

Starting later has an advantage. You can decide what success looks like before you build anything, and pick a business outcome, not a usage number.

That means starting with a single workflow, not a company-wide rollout. Pick the one where a slow or manual step has a visible cost. Write down today’s baseline: time per task, error rate, cost per unit, volume waiting in the queue. Then build the automation and measure the same number again.

That is how we worked with Offerpad. We started with one workflow, vendor photo review, and the result was 1 to 2 hours saved per day. Read the Offerpad story.

This is also why we start with the work, not the model. A workflow with a clear owner, a clear baseline and a clear finish line will show its value in weeks. A tool handed to everyone with no target will show up on the bill long before it shows up in results.

A five-step value check you can run this week

1. List your three slowest manual workflows. Look for repeat work with a visible cost, like invoice questions, intake forms, report building or follow-ups.
2. Pick one. Choose the one with a clear owner and a clear finish line.
3. Set a baseline. Time per task, error rate or cost per unit, measured before AI touches it.
4. Set a target. One person, one number, one date.
5. Run a small pilot and measure again. Keep it if the number moves. If not, change the workflow or drop it.

Once AI is running at scale, vendor reports like Anthropic’s become a useful input. They should not be the scoreboard.

Where to start

If you have not started with AI, the best first step is small: one workflow, one number, one owner.

The Endurance Group is a Claude Certified Services Partner and the outsourced Claude team for your business. We assess, build and manage: a free assessment to find where Claude creates measurable value, a fixed-price build, and Managed Claude to keep it improving.

Sources

• Anthropic just launched tools to show companies what their AI spending is buying, TechCrunch, Oct 28, 2026
• Anthropic reveals AI’s real ROI, The AI Report, Oct 29, 2026
• AI Engineering Trends, October 2026, Jellyfish Research