Most people treat AI like a single tool with one job: do the thing faster. Write the email. Summarize the document. Generate the copy. That framing works for maybe half of what AI can actually do for your business, and it actively works against you for the other half.
There are two distinct modes of working with AI. Automation handles work for you. A thought partner workflow works with you through the process. Knowing which mode to use – and when – is the actual skill. Most companies are building only one of them.
“Start with thought partnership, even if automation is your end goal. Thinking through a process manually first produces automations that actually work.”
Key Takeaways
- AI automation belongs on high-volume, low-stakes tasks where speed matters and a fixable mistake won’t break the business.
- Thought partner workflows belong on strategic, creative, and positioning work where your judgment and experience are the point.
- A thought partner AI asks you questions rather than writing answers for you. That distinction determines whether the output is yours or generic.
- You can use one tool for both modes – you just have to know which mode you’re in and resist collapsing them together.
- Start with thought partnership, even if automation is your end goal. Thinking through a process manually first produces automations that actually work.
AI Automation Belongs on Tasks That Are Already Figured Out
When a task is repetitive, the inputs are predictable, and a mistake is catchable, automation is the right move. Full stop.
The frame I give clients: treat the AI like a team member you just delegated to. The first few runs will surface exceptions and edge cases you didn’t anticipate. That’s normal. You give corrective feedback, the automation gets refined, and over time the error rate drops because the edge cases get resolved one by one. That’s management – not trial and error as a philosophy.
Where companies go sideways is chasing enterprise-scale AI initiatives before identifying the right starting point. The consistent recommendation: find the tasks that are high-volume and not mission-critical, and start there. If something breaks, it can be fixed. That gives your team room to learn how AI performs in your actual environment before you hand it anything that cannot be undone.
According to McKinsey’s 2023 State of AI report, 50% of companies that have adopted AI report cost reductions in at least one business function – but the gains cluster heavily in process-oriented, repetitive workflows rather than strategic decision-making. That pattern holds up in practice: automation built on a well-understood process runs cleanly. Automation built before you understand the process just makes your confusion faster.
The Thought Partner Model Produces Better Output Because It Asks You Hard Questions
A few years into using AI for content, I identified a recurring problem: asking AI to research a topic or draft an article produced output that covered existing ground or missed where my actual thinking lived. Her perspective, her client patterns, her take on why something works – none of that appeared in the output, because it had never been put in.
That’s when she moved to an interview-based workflow. Rather than asking the AI to produce something, she gives it a framework and a role, and the AI interviews her – asking about her perspective, her examples, her reasoning. Then it uses what she’s given it, written in her established brand voice, to build the piece.
For content work specifically, I structure these sessions within the StoryBrand framework – Donald Miller’s approach to narrative and messaging – because it covers the necessary ground without requiring five different systems at once. The AI holds the structure and pulls her through it. What comes out reflects her thinking because the AI never filled in blanks she left empty. It asked until she filled them herself.
**This mechanism showed up most clearly during a training session.I run a class where participants set up their AI coworker, and one section has them work through brand voice and ideal client profile using an AI-led interview. The AI asks questions. Participants answer from their own knowledge and experience. The AI synthesizes.
What kept coming back in that room was a version of the same comment: “It asked me things I hadn’t thought about, but I knew the answer immediately.”
That’s exactly the mechanism. The AI asks the questions a good thinking partner would ask. The person holds the answers already. The output belongs to them because they produced everything in it – they just needed the right questions to surface it.
Ethan Mollick at Wharton has described something adjacent to this, noting that AI-prompted structured reflection produces more authentic, higher-quality work than pure generation. His research at One Useful Thing covers the underlying findings. My experience with clients confirms the direction of that finding, even when the business application looks different from an academic setting.
Thought Partnership Should Come Before Automation, Even When Automation Is the Goal
Here’s the sequencing principle that ties both modes together: use the thought partner first, automate second.
When building a new automation for the first time, I don’t architect it from scratch and hand it to the AI to run. She works through it manually with the AI as a thinking partner – describing the process, letting the AI ask clarifying questions, mapping the logic together before anything gets built. Only after doing it herself, with AI support, does she convert it into a workflow that runs without her.
This matters because automations built on shallow process understanding fail at the edges. The thought partner conversation surfaces those edge cases before they become expensive errors.
Cal Newport has written about the risk of using AI to speed up shallow work without improving what the work actually produces. The thought partner mode addresses that risk directly. The sequence requires thinking better first, then producing.
A 2024 Microsoft and LinkedIn Work Trend Index found that 75% of knowledge workers already use AI at work, but only 39% of companies have provided clear guidance on how to use it strategically. That gap is where the mode confusion lives. Teams default to automation behaviors across all their work because nobody has mapped out when a different approach would serve them better.
Automation vs. Thought Partner: A Side-by-Side Comparison
| Factor | AI Automation | Thought Partner Workflow |
|---|---|---|
| What AI does | Executes the task for you | Interviews and questions you through the task |
| Best for | High-volume, repetitive, low-stakes processes | Strategy, positioning, creative, communication |
| Human role | Delegate and review | Active participant throughout |
| Output ownership | AI-produced, human-reviewed | Yours, structured by AI |
| Risk of going wrong | Edge cases in execution | Sliding toward generation when questions get uncomfortable |
| When to build it | After you’ve done the process yourself once | Before you try to automate anything new |
Frequently Asked Questions
Can one AI tool handle both modes, or do I need different tools for automation versus thought partnership?
One tool can handle both. I use the same AI environment for automations that run without me and for the interview-style content sessions where I’m actively in the conversation. The mode is determined by how you set up the interaction, not by which tool you open. The risk is collapsing them together without realizing it. If the AI starts generating and you stop responding, you’ve drifted from thought partner into autocomplete. The discipline is staying in the mode you intended.
My team has been using AI for months and
it feels productive. How do I know if we’re missing
the thought partner mode?
Look at the work that requires your
experience, judgment, or positioning, and ask whether
AI is involved in it at all. If your team uses AI
heavily for execution tasks and leaves strategy and
communication entirely to human-only drafting, you’re
likely leaving quality on the table. If your team
uses AI to draft strategy documents and
communications without a structured interview process
first, the output probably sounds competent but
generic. Both patterns are incomplete – and the gap
tends to widen as the stakes of the work increase.
What does “building your brand voice in
the AI” actually involve?
Specific inputs: examples of writing you
like, words and phrases you use, words you want to
avoid, how you explain complex ideas, the tone you
want in different contexts. I built mine over time by
having the AI ask me questions about how I speak,
write, and teach – and what feels wrong when AI
produces something that doesn’t sound like me. The AI
jargon list, the sentence patterns to avoid, the
rhythm she prefers – all of that went in. When she
runs an interview-based session now, the output
doesn’t need heavy rewriting because the voice was
established before the content started.
About the Author: Audrey Kerchner is an AI strategy advisor at Inkyma who helps mid-market companies design AI systems that reflect how their people actually work – not how software vendors assume they work. She specializes in building both automation workflows and thought partner frameworks that keep human judgment at the center of high-stakes decisions. Her client work spans content systems, process automation, and AI training programs for business teams.
If you have a process that’s costing your team more time than it should, start there. Use our Process Handoff ROI Simulator to identify where automation would have the fastest payoff – and what to think through before you build it.