Tuesday Email: Processes That Think

Happy Tuesday!

Every Tuesday I'd like to offer strategies for the week ahead and a thought to fuel your action.

I spoke to a peer last week about their meeting prep process.

It’s a three-hour process performed by a CSA.

It’s not about researching investments or doing a financial plan.

It’s about gathering data from five systems and manually inputting the data into a spreadsheet.

Just so they could start running a financial plan.

This firm spends 5% of their revenues on technology. They have the leading solutions. Leveraging automation in their CRM, automating portfolio rebalancing and trading.

Yet, in the middle of their operations they have individuals copying and pasting between systems like it was 2010.

It was in this conversation when I remembered what Nobel Laureate Robert Solow said: "You can see the computer age everywhere but in the productivity statistics."

That was in 1987.

Almost forty years later, we're still living it.

We’ve come remarkably far as an industry; holistic planning platforms, tax optimization software, AI note-takers.

And yet, despite this growth, we remain holding an uncomfortable truth: our margins haven't expanded nearly as fast as our technology budgets.

As we spend more on technology than ever before, we continue to deal with rising cost of human labor.

Teams continue to be overwhelmed, spread thin and drowning in the operational minutia that steals the time technology was supposed to give us back.

Our strategy remains the same: buy more software and hire more people.

Technology has become a necessity, not an efficiency mechanism.

So what’s the problem?

Is the technology failing us, or are we failing the technology?

I think it’s us.

Yes, we are adopting technology. We have a ton of it!

But it’s not about having more, it’s about applying the more the right way.

We keep layering new tools onto old workflows and wondering why nothing changes.

What we desire: more productivity, more bandwidth, more time for clients, better service, the ability to serve more families… none of that happens just with some new technology.

We must rethink how we work, not just what tools we use.

The paradox that we are battling lies in the middle of our firms.

We can break our entire organization into three layers: front office, middle office, back office.

The front office is where the advisor lives and they have seen real gains from technology innovations. Trading is more efficient and prospecting is more streamlined.

We see the same value being derived in the back office. Reconciliation is faster. Billing is more automated. Those processes have drastically improved since we began our business 30 years ago.

But the middle office paints a different picture. This isn’t a performance problem. Your CSAs are not the issue. The systems are. It’s the epicenter of this paradox.

The CSA lives in the middle office. The area where data from one system gets verified by data from another. It’s where we shift from one system to another and back yet to a third system.

It’s where data from multiple systems has to be synthesized before anyone can make a decision.

Best-in-class technology is sprinkled throughout our firms. Planning software, CRM, portfolio management. The bar has been raised on technology and we all have the best in class.

But we continue to struggle with a similar problem: the connective tissue between these systems. That onus continues to fall on our human capital.

Take the best CSA at your firm. Odds are they’ve spent five hours this week pulling reports from three different systems, cross-checking notes against CRM records, and manually building the spreadsheet that lets the advisor actually start planning.

Five hours of prep. For one meeting.

And likely, your firm is tech-forward.

And this CSA is a high achiever and home run employee. They are just dealing with systems that don’t talk to each other in the language that they need them to speak.

The problem is that we have not moved the middle office from gathering and collecting to reviewing and approving.

The automation that we know of and use today in our industry is built on the concept of “if this, then that.”

If a client completes onboarding, send a welcome email.

If a task is marked complete, trigger the next task.

If a cash balance drops below a threshold, generate an alert.

This is known as Robotic Process Automation - RPA. And it has been the foundation of all automation we have been leveraging over the past decade or so.

The thing about RPA is that they are digital bots replicating what we would do, but they aren’t intelligent. They can execute, but they can’t think and learn.

But to execute, RPA needs structured data and deterministic logic (if X happens, do Y - no exceptions). Meaning that it thrives in predictable environments where everything follows the exact path that we program. When that happens, voila. When it doesn’t, well, it’s frustrating.

The thing is that those times it doesn’t work aren’t spurred by large scale changes. It can be as small as a form moving a box to a different section or changing the wording. And the whole thing breaks.

Which isn’t just frustrating, it turns managing the automations into its own job.

As long as the decision to act remains purely human, automation can only compress the time of execution, not the time of cognition.

For instance, automation can help us create a billing spreadsheet automatically, execution. But we are still left to review each row and check for billing errors.

The execution work is removed. But it gets replaced with cognition work. We have traded one burden for another.

What if there was a different path?

The billing spreadsheet is created and potential billing errors are flagged.

This has not just automated the execution but also reduced the cognitive load. We go from looking at 100s of rows to a just a few.

And that is the shift we need to make.

This tension isn’t foreign to me.

It just wasn’t always clear while I was living it.

I’ve always struggled with tools like Zapier. They are great solutions, but so rigid.

It always felt good after you built one Zap, but then you found you needed 15 more to get all the automation you wanted.

And once you have 15 you now need to remember which trigger is starting which Zap and which connection is connected to which trigger and what each trigger outcome produces and where.

It was a plate of spaghetti without a real chef to help understand it.

But here is the irony; I went on to build a company based on the same ethos!

The basis of our company, benjamin, was bringing the same “if this then that” logic but specifically to wealth management.

We connected systems. Automated handoffs. Triggered workflows.

And I felt we had solved the problem.

I remember one demo as I was sitting in my home office on Zoom and I was showing how benjamin could automate onboarding and the prospect had great questions. And I found myself building them a plate of spaghetti.

They asked “what if the prospect forgets to fill out a question?” “What if they don’t fill the form out completely?” “What if they don’t use the portal?”

And my response each time was “you’d need to set up a different automation for that scenario.”

One process now had 15 automations.

It was at that point I realized I was defending our approach, for something that was annoying to me somewhere else.

I remember the first time I used ChatGPT. The ease of having a conversation instead of programming a sequence.

That was incredible.

But it was just the beginning.

The real moment came when I connected Claude to Notion, my knowledge management system.

And I asked Claude a question about a document in that system.

And I got a response, a correct response.

But I then got a suggested update to that document. And a question: “would you like for me to update that for you?”

Of course!

And right away that edit was made in my knowledge management system from my conversation in Claude.

That was the reality we all desire.

Not a plate of spaghetti that we must remember, but rather a conversation (as humans do) that allows me to take the actions I desire.

RPA gave us a feeling that we were making progress, but its rigidity held back the real advancement.

The breakthrough didn’t come when I used Claude within the process I’d already built.

It came when I used Claude to rethink how I worked entirely.

A great way to visualize where we are as an industry with regards to automation and intelligent automation.

Driving.

Manual driving is when we have to do everything in the car; the clutch, the steering, the pedal. And it’s also when the advisor and team are doing everything: gathering the data, entering it into systems, analyzing it, making the decisions.

Cruise control is the ability to set the speed and not have to use the pedal, but still need to steer. For us this is when we adopt a CRM and use the basic workflows like triggering birthday emails, sending calendar invites.

Lane assist was an amazing advancement for driving. No longer do we need to monitor speed or to stay in the lane. But it only knew one thing, going straight at the same speed. Anything else threw it off. That’s the same with RPA for us. It can flag cash levels getting low or minor deviations in our system, but anything out of the ordinary requires human guidance.

Autopilot allowed us to be comfortable that things would adjust if something were to happen. Slow down if something was in the road or someone was slow in front of us. And this is intelligent automation in our business. The ability to see what you are trying to do and adjust. It knows the outcome and works to get you there.

In practice: the systems would notice a client’s cash balance is higher than normal and that interest rates have dropped lately. The system wouldn’t just flag the account, it would generate a recommendation to discuss income alternatives, it would draft talking points and schedule a suggested call time. It sees the context and proposes the path.

And finally we have full self driving. The ability to get from your house to your office without touching a thing. For us this is fully autonomous agents. Completing tasks, end to end, on its own.

Most firms today are operating at lane assist, if then rules that work until something unexpected happens.

But the frontier is moving from lane assist to autopilot. From RPA to intelligent automation. And before long, to autonomous agents.

Here is what this looks like in practice.

Instead of having a recurring task saying “call 10 clients this week” and then having to decide which clients to call and why. Imagine a world where you get an alert that says “call these 3 clients: Jon Smith (daughter finds out about college today), Jane Doe (portfolio is down 5% this month) and Mark Poe (portfolio is down 8% and they haven’t logged into the portal in 45 days).

In the alternate world the system isn’t just reminding you to do something. The system is working with you. And telling you which clients to focus on and why.

This is what automating cognition and not just tasks looks like.

Determining the tasks that need this level of automation and those that need more human in the loop can be a challenge.

I found that looking at this through the lens of the Eisenhower Matrix, with an AI filter applied, can help.

In this reimagined version we would look to categorize tasks and decisions based on two areas: data complexity and outcome stakes.

Low Complexity + Low Stakes: Automate Fully

Tasks like address changes, report creation and generation and scheduling.

These should be fully autonomous with zero human touch.

I have heard many advisors push back on these and say that these tasks are opportunities for a touch point. But I disagree with this.

These aren’t value add. Removing these from your to-do list allows you to focus on touch points that are more valuable and value-added.

Low Complexity + High Stakes: Human-on-the-Loop

Tasks like executing large block trades or processing wire transfers.

With these the data is straightforward, but error consequences are severe.

AI can execute, but a human should be monitoring in real-time and have the ability to overtake the AI immediately.

In this situation, you move from doing to monitoring. From being the doer to being the manager.

High Complexity + Low Stakes: AI-Assisted

Tasks in this quadrant are things like drafting client emails, summarizing meeting notes.

These tasks have unstructured and messy data. But if you get these things wrong, it’s easily fixable.

In these situations, the AI should create the first draft for you. Then you refine it, make it sound like you, ensure accuracy, then send.

High Complexity + High Stakes: Human-in-the-Loop

In this quadrant we find tasks like financial planning recommendations, tax strategy, estate reviews.

The data needed for these actions and tasks is complex and unstructured. And the stakes for getting this output right are high.

This is the area where fiduciary liability lives.

AI can analyze and propose options in this quadrant.

But a human must validate and make the final decision.

Intelligent automation is not a substitute for fiduciary judgment, it’s a synthesizer of the data required to exercise that judgment.

AI acts as the junior analyst, doing research and presenting options.

You remain the senior advisor, exercising judgment.

I’ve found that this version of the Eisenhower matrix gives us the clarity to explore areas to automate, while also reining us in from automating and using AI in areas that may overextend us and create too much risk.

Use the matrix to audit your current processes. Answer questions such as “where are we using humans for robotic work?” or “Where are we lacking intelligent support?”

Much of the pushback in even attempting to experiment with any of these things is that data security and regulation can make this really hard.

And that is true.

Our industry, rightfully so, has real constraints. We are fiduciaries and we are dealing with people’s life savings.

And in a world overwhelmed by technological innovation and cyber crime, data security is (and should be) a top priority.

Regulation is meant to provide us a framework and guardrails. It is not meant to tell us to “not innovate.”

The guardrails are meant to guide us to innovate and evolve within barriers. Innovation doesn’t equate to breaking the rules.

Innovation has always been about understanding what is allowed and not, then pushing the limits of what is allowed.

Intelligent automation and AI is not a substitute for fiduciary judgment. We are still the fiduciary.

AI is a synthesizer of the data required for us to exercise that judgment.

It’s a tool to help us reach the decision point faster. To help surface what we need to see.

But we still make the call.

Regulation is real (and necessary).

But regulation is not the biggest obstacle.

The bigger challenge is us.

The one thing that's stayed constant through every technology transition in our industry is how we work.

We've applied innovations to existing workflows. New tool, same process. Better software, same workflow.

Just having efficient tools doesn't make the firm efficient.

A process and structure and strategy for having those tools work together is what provides efficiency ROI.

We've been spending money and adopting technology, but we haven't been creating a system around it.

We haven't been rethinking how we work.

This isn’t design it once and you’re done.

It’s design, then do, then train, then do, then retrain, then do again.

30-year veteran pilots don’t stop training. They go through recurring training because it is a high stakes environment. Airlines and pilots can’t afford to have their skills decay.

Our industry is high-stakes too. Ongoing practice isn’t optional.

The transition from automating tasks to automating decisions isn't about replacing the advisor.

It's about elevating the advisor.

Trust remains the currency of the realm.

Clients aren't looking for a robot.

They're looking for you, a deeper, more meaningful connection with you, the human.

The path forward requires keeping the human at the helm while letting AI power the engine.

In 1987, Solow couldn't see productivity gains in the statistics.

In 2026, we know why.

We've been measuring the wrong thing. We've been automating tasks when we should have been redesigning work.

The winners of the next decade won't be those with the most technology.

They will be those who rethink how work actually gets done.

The best is ahead!

-Matt