Essay · Markets & enterprise
The $2 billion is the size of the problem
LPL told 6,000 advisors it has put nearly $2 billion into technology. The industry read that as a commitment. It is a receipt — and at a firm with 32,000 advisors, most of a number that size never reaches anything new.
Jason Wenk asked the obvious question under Robert Noe's post about LPL's announcement this week. “Is this sorta the same thing as saying your platform is so bad today it needs a $2b overhaul?”
I said what I think:
Agreed, although I think the more tech and operating debt a business has the more money dealing with the problems it will take and the lower chance there is of success.
Incumbents in our industry are going to fall in the next decade.
Here is the case for that.
The money is already spent
LPL told about 6,000 advisors in San Diego that it has put nearly $2 billion into technology. The number moved through the industry in a day, and it moved as a promise. Noe's post called it a “$2B commitment in rebuilding their technology stack” that “will fund Latitude.”
That is not what LPL said. LPL said it has invested nearly $2 billion over the last three years. Past tense. The money is gone. Latitude is what it bought.
Almost nobody carried the distinction. A commitment is a plan you can hold a firm to. A receipt is a record of what a firm needed to spend to get where it is now.
What it bought
Latitude packages five things: LPL's data architecture, its infrastructure and the Business Browser it deployed earlier this year, ClientWorks, Account View, and an AI agent called Cyan. Four of those are live and were live before July 28. Cyan is the one everyone wrote about. Cyan launches later in 2026.
Nothing new shipped at Focus.
For timing: Cetera put IntelligenceEngine in front of its advisors in July 2025. Raymond James launched an internal AI agent in January 2026 and took Client 360 out of beta in May. Altruist shipped Hazel in September 2025 and has been adding an agent a quarter since. Osaic did not build anything — it bought Jump and Zocks and got roughly 30% adoption inside a year.
LPL announced in July 2026 and will ship later.
The 2018 sentence
ClientWorks was unveiled at LPL's Focus conference in August 2014 as the next-generation platform, with full release promised for spring 2015. In March 2018 — three years past that date — LPL delayed retiring BranchNet again. Advisors quoted at the time wanted ClientWorks “cleaned up and bug-free” before leaving a system they had used for seventeen years.
And LPL's spokesman said this:
We plan to go live with more than 40 new enhancements to the platform in the next 60 days.
The 2026 line is “more than 35 major technology enhancements” this year.
Donald Windle said it in six words in the same comment thread: “They just ‘rebuilt it’ in 2018.”
He is right. It is the base rate.
The number does not reconcile
LPL's own second-quarter investor presentation carries a line called Core Technology Portfolio Spend. For 2023, 2024 and 2025 it reads roughly $310 million, $265 million and $375 million. That is about $950 million — about half the announced figure. A second line, capabilities for new institutions and M&A, adds roughly $440 million across the same three years, and that is acquisition integration rather than platform.
So about $1.4 billion of disclosed spend against a stated $2 billion, with no public reconciliation. The rest is likely capitalized software, cybersecurity, and keeping the existing estate running.
Which is the point. At a firm with 32,000 advisors, most of $2 billion over three years is not a rebuild. It is maintenance and defense. The bigger the installed base, the larger the share of any technology budget that never reaches anything new.
Two hard things in the same quarter
Cyan launches later in 2026. Commonwealth's roughly 2,900 advisors finish moving onto ClientWorks in the fourth quarter of 2026.
Those are the same quarter.
LPL paid about $2.7 billion for Commonwealth and told the market it expected to retain 90% of assets. Asset retention is in the mid-80s. On headcount it is worse — 653 advisors left between April and December of last year, more than a fifth of the firm, and a third of them went to Raymond James.
Moving 2,900 advisors from one platform to another is among the hardest things an enterprise attempts. Schwab did it with TD Ameritrade and spent the better part of two years explaining its service levels to advisors.
Most who moved from TD to Schwab were not so happy with it.
Scheduling that against the launch of your first AI agent tells you where the engineering hours are going.
What AI is coming for at LPL
Noe's post argued that AI is coming for the advisor who was never advising. I would put it somewhere else.
In the second quarter LPL earned about $443 million on roughly $57 billion of client cash. On the same call Steinmeier said the firm is “doing the work to properly assess the opportunities and risks of reducing our reliance on cash sweep economics.” Financial Planning reported it under the heading of an AI threat.
That is the exposure. The spread.
An advisor who has spent a career being told the platform is the value should look at how much of the platform's economics comes from client cash, and at the firm now studying how to depend on it less.
Adding AI onto antiquated rails
Most of what has been written since July 28 is about whether Cyan will be good. My assumption:
Adding AI onto antiquated rails isn't likely to work well.
Putting wealth management services on AI Rails allows you to serve more customers in improved ways with less people. Intelligence with integrated customer specific data doing all the work that doesn't require judgment and organizing and prompting the work that requires expert human judgement.
That is what the rails are for. The data architecture, ClientWorks, Account View and the Business Browser are LPL's. Cyan sits on them. Four of the five parts of Latitude are the thing the fifth part has to work through.
Here is how I have been saying it about every firm still building the old way, incumbent or not:
Pre AI, like the generation what they were creating makes sense. But post agentic AI, they're orienting in the wrong direction with what they're building… A bunch of softwares don't connect to each other very well… Like why would you keep building that? Scrap that and create what ought to exist.
And the rest of what I said:
they already have operating and technology debt in… our industry. And so it's like, gosh, I would scrap what they've done on that side and start over. And anybody older than that, it's way worse.
It is harder the larger you are. It is also the reason the announcement reads the way it does:
LPL has to at least portray outwardly that they are keeping up with tech - even though they have so much tech and operating debt that it's more likely to be a veneer than a strong digital solution.
Transition assistance is the mechanism. A note has a term. A platform that arrives three years late against a note that comes due in three years is not a technology question, it is an underwriting question — and Steven Woods said it more bluntly in the same thread: “$20mmm for the build … $1.98B in transition assistance to not leave for something better for 3yrs while they build it.”
What I did not have to do
Building on agentic rails from the start meant never having to decide whether a capability works for 32,000 different practices before shipping it to one. It is the difference between announcing a capability and shipping one.
There is a version of this I have said more plainly:
There needs to be somebody who actually understands this industry building what ought to exist.
What I think happens
None of this is an argument that LPL is in trouble. It recruited about $25 billion of assets in the second quarter, its best in roughly two years. Retention is 97%. Adjusted pre-tax margin is 39.3%, and it has lowered its expense guidance twice this year. The firm has money and is spending it.
Wenk's read is that these transformations take five or more years, so the sooner you start the sooner you finish. That is the generous version and it may be right.
Mine is narrower. Tech debt is not a fixed cost you pay down on a schedule. It compounds against you, because every year the thing you are trying to reach moves and the thing you are carrying gets heavier. A firm with 32,000 advisors, a 2,900-advisor migration in flight, and a revenue line it is publicly reconsidering does not get to spend its way to a standing start. Spending is what the debt requires. It is not what closes the gap.
Incumbents in our industry are going to fall in the next decade. Not all of them, and not for lack of money. They will fall because the distance between what they carry and what a firm built on agentic rails carries gets wider every quarter.
Embracing and building on AI Rails is a no-brainer. Anyone not building right now has their head in the sand or has limited motivation to improve their and their clients experiences and results.
Seems like most have their heads in the sand.
Sources: LPL Financial Focus 2026 and the Latitude announcement, 28 July 2026 · LPL Financial second-quarter 2026 investor presentation and earnings call — Core Technology Portfolio Spend, client cash revenue, recruited assets and margin · InvestmentNews reporting on the ClientWorks and BranchNet delay, March 2018 · Financial Planning on cash sweep economics · the comment thread on Robert Noe, Jr.'s LinkedIn post, August 2026. Every indented passage is Bill's own words, unedited — from that thread, from his notes on the draft, and from two conversations recorded on 19 August 2026.