Systems Decision Map (TWT)

Hey, I'm going to walk you through the systems decision map skill that you can use inside of ChatGPT or in Claude and why I created this in the first place.

So the systems decision Map, a lot of people ask me often where should they start with the problem, how should they apply AI to their work?

people are, a lot of us are quick to just say like, okay, let's just use AI for it when we don't really know what the real problem is.

And not just the surface level issue, but the root cause of the pain that we're experiencing in our work.

And so let's just walk through this example and we'll show you what this looks like.

So inside of ChatGPT, I'm going, to look for the systems decision map.

So I have a bunch of skills in here and then what I'm going to do is I'm just going to start blabbing, using voice dictation, about the process that I want to have improved.

Hey, so I am working for an accounting firm.

My current proposal process takes four hours and I really like to get it down to one.

Now I am intentionally leaving this very vague and to see how the AI will handle the situation.

Now I do recommend using a higher model, a smarter model like Opus Fable if you're using Claude or with ChatGPT if you're using Sol or Astra.

use a thinking model to help us work through this problem together.

So now what it's doing is it's reading the system decision map skill to understand where the four hours is going to, because where that time is spent will help inform how we should approach a solution.

So it's doing the right thing here.

It's asking where we're actually spending that time.

Are those four hours hands on work, total turnaround, including waiting time for information or approvals?

All right, so I'm going to give it a little bit more context.

We already have AI creating the first version of the proposal from a transcript.

The hard part is determining how to scope and price the proposal based off of the information that we have.

Because we're collecting a lot of information from our clients.

It could be, information about their accounts receivable process, could be information about their business, about, where they're heading, what growth looks like, etc.

And time is, a lot of my time is spent just to, getting the approvals from managers or directors who have been pricing these proposals for a very long time.

I Feel like I'm missing information.

Okay.

So this is a pretty common scenario in value based pricing.

If you have not been selling for a company for very long, you might not have all the details you need or might be uncertain about exactly how to price something because you don't want to underprice or overpriced.

so let's see how it handles this.

Okay.

And this was pretty obvious but it said time is going to scope, pricing and approval.

The outcome that we want is scope and price.

You can get approved with fewer rounds of questions.

That is correct.

gaps.

So client facts that affect the actual work or the decision rules that experienced managers use to turn those facts into scope and price.

And that's exactly right.

This is our issue.

So now it's going to continue to go through and ask questions.

Hey, this is for testing purposes.

Can you answer all the questions that you would normally ask just so that we can see the outcome that's provided by the skill that's being run?

Okay, great.

Let's take a look at what it came up with.

So it is now giving us an idea of like the kinds of questions that it's going to ask because it was missing information.

Right.

Like what are the four hours measure?

What did the last manager need to make a decision?

How does the firm currently price work?

Where are those rules recorded, if at all?

Right.

Mostly in past proposals or managers experience.

There's no shared decision guidelines.

This is pretty common.

What evidence is available, how repeatable are proposals, who has authority and what is the investment budget?

So gets.

So that would be through conversation.

This AI would ask you all of this information based off of the initial information that you provided.

And then it's going to give you something like this.

This is designed to give you six different outputs based off of first principles and second order effects.

First principles to, gets at the root of the problem.

So the outcome is proposal with a clear scope and approved price produced with less staff effort.

One hour is the performance target.

And I really like how this breaks down.

Right.

Client what this requires client facts that explain the work that the firm would take on.

Clear responsibilities, exclusion and treatment of uncertain work.

And then pricing the pricing that the firm actually accepts with an accountable person for approving those commitments.

Okay.

And so it clearly identifies what the constraint of our problem is.

In this fictional case.

The constraint is access to pricing judgment.

All that information lives inside of one person's head or a couple of people's heads.

And we need to extract that somehow so that we can share that judgment across the firm, make it easier to price deals.

Intervention.

This is where it gets a little bit interesting.

Based off of the information it knows now we can say, okay, here are some paths forward that we can take.

Right?

Path one.

This is where we start with a decision worksheet.

This would be shared with both humans and or AI.

Both people are going to need this to make better scoping decisions.

What changes would be, we're going to collect information about client facts and scope choices, pricing rules and unresolved questions.

What this requires would be senior time to explain and agree that those rules are the rules.

And the idea being this could repeat repeated discussions, between all the sales folks and the manager, but the exceptions would still need judgment.

And then how AI could fit into this.

So then it gives you, proposed handoffs as well.

Who's responsible for what in this process and how might we fix this?

And then anytime you're applying fixes to something, so the second order effects, is, if this new system that you've designed actually works as intended, or maybe it doesn't.

There are things that we want to watch out for.

If this works well, what might happen?

Staff may prepare acceptable proposals with less dependence on particular managers, which is great.

Less questions.

Senior staff could spend more time on the unusual and edge cases rather than answering every question.

That's the ideal state, I think.

Here's the important part though.

It also captures watch outs, gotchas, consistency.

And in the system could also hide any mistakes, right?

Like an incomplete rule might cause repeated underpricing, for example, or promises that maybe the delivery team actually couldn't meet.

And so we just want to make sure that we're reviewing these rules over time so that they continue to be accurate.

And then here's another possible effect, right?

If we send out proposals more quickly, we close more deals that could increase sign work, which means that we would need to just keep a close eye on the onboarding team and delivery team and ensure that they actually have the capacity if we are closing more deals.

And then you have things like feedback loops, any new constraints that might be introduced as a result of this system that you're implementing.

So if pricing judgment becomes easier to access.

Missing client facts may become the next limit.

So what are things are we not already accounting for?

And that's it?

That helps you determine what the real problem is?

What are some possible solutions that you can implement right away?

Small steps that you can take whether or not you want to use AI or need AI.

And I kept this short intentionally just so you weren't bombarded with a bunch of information.

But you can always ask to expand on this design and this worksheet and continue to build from there.