Strategy & transformation
2:35 video · 8 min read
The short answer
Watch: AI transformation beyond productivity
Watch the explanation, then explore the practical implications below.
What is strategic AI transformation?
Strategic AI transformation means changing how a business achieves its goals with AI. It connects the technology to decisions about processes, people, information and accountability. The starting question is what the business needs to accomplish, followed by how AI and people can do that work together.
Saving time and reducing costs are valuable outcomes. The strategic question is what happens next. Does the released capacity help the business respond to customers sooner, protect margin, improve decisions or deliver more work with the same team? Leadership needs to make that connection explicit.
Why faster tasks do not always improve the business
A person can complete their part of a process faster while the overall result barely changes. An AI-generated report may still wait a week for a meeting. A quote may be drafted quickly but sit in the same approval queue. A team may create more content while its customers receive no clearer answers.
These are signs that the constraint may sit elsewhere in the workflow. Buying more tools or asking people to produce more can increase the volume passing through the same bottleneck. The useful measure is whether work reaches a better outcome with less delay, rework or risk.
How CSM would approach an AI transformation
1. Start with a business outcome and a baseline
We would begin with a specific process and the result it needs to deliver. For quoting, that could mean a shorter enquiry-to-quote cycle while preserving pricing accuracy and margin. For reporting, it could mean resolving exceptions sooner. Record the current volume, elapsed time, hands-on effort, rework and points where work waits.
2. Redesign the whole workflow
Map the process from the initial request through to the completed outcome. Identify what information is needed, where it lives, who makes each decision and which approvals are necessary. Remove redundant steps before automating them. This is also where individual use of ChatGPT or Copilot can become a shared company capability: agreed knowledge, connected processes and clear responsibilities.
3. Give AI and people explicit responsibilities
Define the work AI can prepare, the actions it may take and the decisions that require a person. An agent might assemble source material, flag missing requirements or draft an answer. A named person should own consequential decisions and exceptions. Access should be limited to the information and actions the task requires.
4. Change measures and decide how capacity will be used
Measure the outcome alongside speed. For a quoting workflow, useful measures might include total response time, correction rates and margin exceptions. Include the time people spend checking and fixing AI output when estimating net savings. Agree which work can stop and whether the recovered capacity should support customers, increase delivery capacity or improve quality.
5. Test, learn and assign ongoing ownership
Start with a bounded workflow and compare it with the baseline. Test ordinary cases, incomplete information and exceptions. Review errors and record approved corrections in shared guidance that the system can retrieve. Decide who maintains that guidance, reviews performance and pauses or changes the process when it stops meeting the agreed standard.
A practical example: improving the quoting process
Consider a company that receives enquiries across emails and attachments. AI could draft a quote faster, but the team would still lose time if requirements were incomplete, prices were hard to find or every quote needed the same senior approval.
A redesigned workflow could bring the enquiry and attachments together, identify missing requirements, retrieve approved pricing information and prepare a draft with its sources. The responsible person would check scope, assumptions and exceptions before approving the quote. The approved information could then support the delivery handover.
This illustrative workflow should be tested against response time, accuracy, review effort, margin protection and the quality of the handover. A faster first draft is only one part of that assessment.
What meaningful human oversight looks like
“A human checks it” is not a complete control. The reviewer needs to know what to check, see the underlying evidence, have enough time and expertise, and be able to reject or change the proposed action. Otherwise, review can become a routine approval that adds effort without catching the important mistakes.
The level of review should reflect the consequences of an error. Formatting an internal note and approving a price commitment deserve different controls. Define escalation rules for uncertainty, missing information and unusual requests, and retain a record of significant decisions.
How do you know whether the transformation is working?
Look for evidence across four areas: the business outcome, the performance of the full process, the net capacity returned and the quality of decisions. Activity measures such as the number of prompts, drafts or reports can help explain adoption, but they do not establish business value on their own.
The test is whether the business can do something useful better, with a clear understanding of the effort and risks involved. If the team is producing more and reviewing more without improving the result, revisit the workflow, its measures and the role AI has been given.
Questions for your leadership team
What outcome are we trying to improve?
Name the customer or business result, the baseline and the person accountable for it. Would producing twice as many reports, quotes or documents actually improve that result?
Where does the work really get stuck?
Is the constraint the task itself, missing information, an approval queue or a handover? Which step could make faster work elsewhere irrelevant?
What will people be able to stop doing?
Identify the activity that can disappear and agree how the returned capacity will be used. Are targets and incentives consistent with that decision?
Can the reviewer challenge the AI?
What evidence will the reviewer see? Do they have the expertise, time and authority to intervene? What happens when the system is uncertain or wrong?
What would make us change course?
Agree the results that would justify expanding, redesigning or stopping the workflow. Include quality, review effort, adoption and operational risk alongside time and cost.
Full video transcript
The full video transcript is included below for reference.
Productivity gains need a different process
I think it’s actually pretty urgent to do a few things. One is recognising that viewing this purely as a productivity tool gets you into trouble in a lot of different ways.
People are getting returns from AI. The productivity gains are real. The problem is that those gains get eaten up by the organisation if you don’t change the process around them.
And I think part of that is we’re asking people to work differently without changing what we expect from them. We give someone AI, tell them to experiment, but they still have the same targets, the same approval process, the same things they’ll be judged on at the end of the month. So, naturally, they use it to get through those things faster.
Then leadership looks at that and wonders why the business hasn’t transformed. But where did you actually give people room to change how the work gets done?
Measure the outcome the work is meant to achieve
If you don’t change your KPIs, people produce more of whatever you’re already measuring. More PowerPoints, more reports, more code. You need to ask what those things were supposed to accomplish in the first place. Was the goal to produce the report, or to make a better decision?
Decide what people can stop doing
So leadership needs to decide what this thing is for. And part of that decision has to be what people can stop doing. If AI saves someone five hours, but they have to fill those five hours with more of the same work to demonstrate that they’re busy, you’ve already decided how that capacity gets used. You just haven’t made that decision consciously.
Make human judgement meaningful
The second issue is that AI is getting very capable. If our default is automation, we just keep handing tasks to AI without thinking hard enough about where humans belong in the process.
And that takes more thought than saying, “We’ll have a human check it.” What are they checking? Do they have the time and the expertise to challenge it? Can they actually change the decision?
Those are choices leadership has to make. Otherwise, you can end up with people producing more, checking more, and feeling busier than before, while nobody is quite sure whether the business is doing better.
Start with one process that matters
Bring a workflow where delays, repeated work or unclear decisions are limiting the business. We can explore the outcome you want, what needs to change and where AI could help.