A lot of companies have some version of the same objective sitting in a strategic plan:
We need to deploy AI.
That sounds a little like me saying, “I bought an automatic leash. Therefore, Dusty and Tanner are walked.”
No. I bought a piece of technology.
The dogs still need a destination. Someone still needs to choose the route. Someone still needs to notice when Tanner sees a bicycle and decides we are all going in a different direction.
Technology is not the outcome.
This is the final question in the Reframe stage of my G-R-O-W-T-H Framework: What result are we trying to create, and what would we design differently if the result mattered more than the technology?
The question is especially important now because many businesses are measuring AI activity instead of AI impact. They count licenses, users, prompts, agents, pilots, and automations. Those measures may tell us whether people are touching the technology. They do not tell us whether the business is growing faster, converting more customers, retaining more revenue, or operating more effectively.
The better starting point is not, “Where can we use AI?”
It is, “Which growth outcome could AI materially improve?”
Start With the Outcome, Not the AI Use Case
When a company starts with the technology, the conversation usually produces a list of possible applications.
We could create an AI chatbot.
We could score accounts.
We could automate content.
We could build a customer-success agent.
We could summarize sales calls.
Some of those ideas may be useful. But they are still descriptions of activity. They do not establish what the business expects to improve.
Reframe the question around a measurable result.
- 🐕 Instead of: Where can we deploy AI in sales?
- 🐕 Ask: Where is seller time being lost, and which delay is slowing pipeline?
- 🐕 Instead of: Should we add an AI chatbot?
- 🐕 Ask: Which visitors fail to find the information they need, and what conversion improvement would justify the investment?
- 🐕 Instead of: Should customer success use an AI agent?
- 🐕 Ask: Which customer behaviors predict underuse, risk, or expansion, and what intervention could change the result?
Once the outcome is clear, AI receives a job. Without the outcome, AI becomes another tool the organization is expected to adopt without knowing what success means.
Example 1: Reduce the Time From Account Signal to Seller Action
Imagine a company with thousands of target accounts.
Its sellers have access to CRM history, intent signals, website engagement, product usage, company news, and past opportunity data. The information exists, but it is scattered. Sellers spend hours deciding which accounts deserve attention and why.
An AI system could combine those signals and prioritize the accounts most likely to warrant action.
But “deploy AI account scoring” is not a sufficient objective.
A measurable growth outcome would be:
Reduce the time between a meaningful account signal and seller action from five days to five minutes.
That statement changes the design conversation.
Now the team must define what qualifies as a meaningful signal. It must determine which data the AI can access, how the system ranks or interprets the signal, what recommendation reaches the seller, and what action should follow.
It also creates a clear test. Did response time improve? Did sellers act on the recommendations? Did faster action create more qualified conversations, opportunities, or pipeline movement?
The technology becomes part of an operating system connected to a business result.
Example 2: Give the AI Chatbot a Conversion Job
Consider a website serving multiple industries and buying roles.
A CIO, operations leader, technical evaluator, business user, and procurement stakeholder may all arrive with different questions. A generic experience forces each visitor to search through the same navigation, pages, and resources.
Leadership may decide to add an AI chatbot.
Again, that is a feature decision, not an outcome.
A stronger objective would be:
Increase qualified visitor-to-meeting conversion from 2 percent to 3 percent.
Now the chatbot has a specific job. It must recognize or discover the visitor’s context, provide useful answers, guide the visitor to relevant proof, and create an appropriate next step.
The team can monitor whether qualified visitors engage, whether they receive relevant information, whether meetings increase, and whether those meetings become real opportunities.
If conversion does not improve, the business should not declare success because AI adoption was high. Usage may be a leading measure, but it is not the result the system was built to create.
Example 3: Use AI to Improve Customer Adoption and Retention
The same discipline applies after the sale.
Suppose customers who adopt a particular set of features are more likely to remain, expand, or achieve value. Other customers underuse those features and become vulnerable to churn.
AI could identify usage patterns, flag at-risk accounts, recommend the next best action, create customer-specific education, and trigger outreach.
The weak objective is:
“Deploy an AI customer-success agent.”
The stronger objective is:
Increase adoption among at-risk accounts by 15 percent and reduce churn.
Now the organization knows what the system is supposed to influence. Customer success can evaluate whether the recommendations were relevant, whether customers took the intended action, whether feature adoption changed, and whether retention improved.
Subscribe for more resources on this topic here
Define the Human Role Before the AI Starts Working
AI does not automatically replace the marketer, seller, or customer-success manager. It changes the work each person performs and the decisions that deserve their attention.
When I walk Dusty and Tanner, I could use a device that recommends the best route based on weather, distance, pace, traffic, or the presence of other dogs.
That information would be useful.
But if Tanner suddenly sees a bicycle and bolts, I still want judgment attached to the leash.
Businesses need the same design discipline. Before an AI system enters a growth workflow, answer four questions.
- 🐕 What can AI decide? These are decisions the system can make within clearly defined boundaries, using reliable data and agreed rules.
- 🐕 What can AI recommend? These are situations where the system can identify patterns, rank options, or suggest a next action, but a person retains responsibility for the decision.
- 🐕 What can AI execute? These are repeatable actions the system can perform after the required conditions, permissions, and safeguards are satisfied.
- 🐕 Where must a human approve or intervene? These are decisions involving strategic judgment, material risk, sensitive communication, exceptions, or consequences that should not be delegated blindly.
The goal is not to keep a human in every step. That would eliminate much of the value. The goal is to put human judgment where it matters and let AI handle the work it can perform reliably.
Move From AI Features to Results as a Service
This outcome-first approach points toward a larger shift.
Customers do not ultimately want software for its own sake. They want the result the software helps them achieve.
They want:
- 🐕 Fewer outages
- 🐕 Faster hiring
- 🐕 More revenue
- 🐕 Lower risk
- 🐕 Higher productivity
- 🐕 Better customer retention
As AI systems become more capable, companies can increasingly organize their offerings around delivering those outcomes instead of giving customers another tool to configure, operate, and monitor.
That is the idea behind results as a service.
Software gives the customer a capability. Results as a service makes the desired outcome central to the product design, operating model, customer experience, and measurement system.
That does not mean every company can guarantee an outcome it does not fully control. It means the business starts with the customer result and designs the combination of AI, data, people, process, and governance required to move it.
Apply the Same Principle Inside the Company
The same shift should happen internally.
Do not give marketing an AI tool and call it transformation. Give marketing an outcome.
Do not give sales an AI assistant and call it productivity. Define the change in response time, account coverage, conversion, pipeline velocity, or seller capacity that the assistant is expected to create.
Do not give customer success an AI agent and assume retention will improve. Define which behaviors the system should detect, what intervention should occur, and how customer adoption or churn should change.
For every AI initiative, connect five elements:
- 🐕 Outcome: What business result should improve?
- 🐕 Baseline: Where are we starting today?
- 🐕 Target: What measurable change are we trying to produce?
- 🐕 Operating design: How will AI, data, people, and process work together?
- 🐕 Accountability: Who owns the result and decides whether the system is working?
This is the difference between installing technology and redesigning work.
A Practical AI Outcome Brief
Before approving an AI pilot, require the team to complete a one-page outcome brief.
1. Business problem
Describe the delay, friction, cost, risk, missed conversion, or customer problem in plain language.
2. Baseline
Document what happens today. How long does the process take? What percentage converts? How often does the problem occur? Which customers or employees are affected?
3. Target outcome
State the expected improvement as a number, time reduction, quality threshold, or observable business change.
4. AI responsibility
Define what the system will detect, recommend, decide, generate, or execute.
5. Human responsibility
Define where a person reviews, approves, overrides, handles exceptions, or remains accountable.
6. Required data and systems
Identify the data sources, integrations, permissions, and workflow connections required for the system to operate.
7. Success measures
Separate adoption metrics from outcome metrics.
Adoption metrics might include usage, active users, recommendations viewed, or actions completed. Outcome metrics should show whether the business result improved.
8. Stop, revise, or scale decision
Set the conditions that determine whether the initiative should expand, change, or end.
This brief makes it difficult to hide behind activity. The AI either contributes to the intended result or it does not.
The Final Reframe Question
Reframe began by turning the map around.
Instead of reacting to today’s problems, leadership defined where the company wanted to go. It selected the outcomes that matter, established a North Star, reconsidered the ideal customer, chose the right go-to-market motions, tested whether the necessary capabilities existed, and examined how teams should work together.
AI belongs at the end of that sequence for a reason.
Technology should support the strategy. It should not become a substitute for one.
So before the company announces another AI initiative, walk the idea around the block.
What result are we trying to create?
What is the baseline?
What should change?
What can AI decide, recommend, or execute?
Where must human judgment remain attached to the leash?
Who owns the outcome?
And what would we design differently if the result mattered more than the technology?
That brings us home.
Gauge told us where we were. Reframe forced us to decide where we are going, who we want to win, how we plan to grow, what capabilities we need, how teams work together, and what outcomes technology should deliver.
Dog walked.
Now we are ready for O: Orchestrate, where all those choices have to work together in the real world.
