Bloomberg Podcasts Sep 20, 2026
With Jenny Johnson, CEO of Franklin Templeton
Franklin Templeton's internal AI tool has its salespeople in front of 23% more clients and has lifted sales 11%, and the chief executive still says the rest of what the firm is spending on AI has not paid off.
The public argument this month has been whether the companies building AI should slow down. Jenny Johnson's answer is that the question does not reach the people buying it: a chief executive with today's models in hand has years of work left before pacing is even an option.
"You can't pace yourself."
Johnson ran Franklin Templeton's technology department before she ran Franklin Templeton, which is where her two most specific claims come from — that an application with a big installed base is almost impossible to rip out, and that the thing companies are about to get wrong is the governance layer over their own agents.
The full segment is covered here so you can skip it.
Here are the 7 lessons that matter.
Key Takeaways
A CEO cannot pace themselves even if the labs do, because the models already shipped will take years to exploit Her second reason is competitive: if the US slows and China does not, the lead goes
Programming has inverted — the job is now to specify everything the model must not do, rather than what it should
The governance layer is what nobody is building, and the companies selling the models have no reason to suggest it Her example is an agent left running daily, unwatched, burning tokens
Franklin Templeton's own sales tool has its people seeing "23% more clients" and produced an "11% uplift in sales"
Everything else the firm is doing with AI "hasn't paid off yet", and she says so while insisting it eventually will
AI has not rerated Franklin Templeton's margin, and the AI companies themselves went from capital-light to capital-intensive with the expense not yet through the P&L
The Fed had to raise rates to stay credible, and she reads the market's reaction as approval of Kevin Warsh's independence rather than of the decision
1. You Cannot Pace Yourself
The segment opened on the week's AI anxiety and the calls from AI company chief executives for the industry to slow itself down. Johnson said the fear is legitimate and then split the problem in two: what frightens her as a citizen, and what she has to do as a chief executive.
The behavior that worries her is models breaking into environments they were not given
You have these models that are going out, and not only are they breaching other people's environments
Jenny Johnson
The second bucket is the operating one, and it is where she spent the rest of the answer. Asked whether the slow-down argument was therefore beside the point, she cut in before the question finished.
Pacing is not available to a company that has not caught up yet
You can't pace yourself.
Jenny Johnson
The models already shipped will take years to exhaust
it is gonna take us years before we maximize the existing kind of front end models
Jenny Johnson
Having run technology herself, she described the change in what writing software now means. The instruction set has flipped from the permitted to the forbidden, which is a harder specification problem because it has no natural end.
The job is now to enumerate what the model must not do
When you programmed, you programmed to tell it what to do. What's happening is you almost need to program to think about all the things you tell it it can't do.
Jenny Johnson
Her second argument against slowing down is competitive rather than commercial, and she stated it as a fact about where the two countries stand.
A US slowdown that China does not match hands over the lead
If we slow down and China doesn't slow down, we sort of see that leadership position. We can't afford to do that either.
Jenny Johnson
The third argument is that bad actors will not observe the pause
in life, there's always another Putin, Hitler, some megalomaniac who's gonna act badly. You can't give them the tools and not have any offsetting side.
Jenny Johnson
She closed the answer where she had started it, on how little of the available capability is in use.
Most companies have barely begun
we're only scratching the surface of leveraging these models within our own environment
Jenny Johnson
2. The Governance Layer
Asked whether the margins work yet — whether a company is getting back what it pays for agent-based AI — Johnson did not answer on price. She answered on control, and named the missing piece.
The thing that is about to matter is governance, not capability
what is gonna become very important is the governance layer
Jenny Johnson
The sequence she described is deliberate and has a cost attached. Putting the models in everyone's hands first is what makes people comfortable with them; the bill arrives later, when those people start building agents that nobody is watching.
An agent with no owner runs every day and spends every day
the risk you have is that people start generating these agents. And, oh, by the way, it's just running every day, and nobody's looking at it.
Jenny Johnson
The anchor's contribution was a joke about a man whose automated away-message was going to cost him $10,000, which Johnson took as the example rather than the punchline.
The second half of governance is routing — deciding which model a query is worth.
Most queries should be sent to the free model
And then the second piece is a lot of what you do should go into the free model. So you need the routing to say, oh, this is not a difficult query. This is an easy agent. Let me send it over to the free one, or this is a really challenging one.
Jenny Johnson
She then named the conflict of interest, without softening it.
The model vendors have no reason to teach customers to spend less
big AI companies aren't encouraging you to think about that governance, but I think that discipline is gonna be really important.
Jenny Johnson
3. 23% More Clients
Johnson gave one worked example of AI paying for itself inside her own firm, and it is a sales-productivity tool rather than anything customer-facing.
The intelligence hub decides which clients the sales force should be in front of
We call it the intelligence hub. It's essentially making our salespeople's times more effective, visiting the right clients, having the right conversations.
Jenny Johnson
Two numbers came with it
they're seeing 23% more clients because it's taking away a lot of the administrative work, and we have 11% uplift in sales
Jenny Johnson
The candid part came immediately after, unprompted: the rest of the program has not returned anything yet.
Everything else is still unpaid, and she expects it to take time
Some of the other stuff we're doing, we're still kinda looking at it and saying, okay. It hasn't paid off yet, but I think it'll take time, and there's no question in my mind that it will pay off.
Jenny Johnson
4. The Installed Base
The anchor put the software sector to her: a group that spent the "SaaS apocalypse" worried about its own existence, then reported strong earnings and led the technology sector higher over the previous week. What did she make of that?
Johnson granted the threat first.
Handing programming to non-programmers is a real risk to software vendors
For sure, you are putting in the hands of the average person the ability to programming, and that potentially has a threat.
Jenny Johnson
Her evidence is a family one, and it is a company small enough to walk away from a subscription.
Her son-in-law's cleaning business wrote its own CRM rather than buy one
My son-in-law has a little cleaning company. They've written their own CRM system. Right? So they're no longer going to Salesforce to go write it. They're doing their own.
Jenny Johnson
Then the counter, from having run a technology department: the defense is not the software, it is everything downstream of it.
An application with a big installed base is almost impossible to remove
once you have an application that's a big installed base, it is feeding this downstream all your other systems. It is so hard to pull it out.
Jenny Johnson
The anchor offered his own supporting statistic, which is the show's rather than hers: probably half of financial services firms are still running a mainframe in their back office. Johnson agreed, added that they might not admit it, and extended the point to the person running a business on an oversized Excel grid.
Her conclusion separates the revenue at risk from the multiple at risk. The incumbents have time; what they may not keep is the rating.
The risk is to the multiple, and the defectors are the small accounts
The question is, does their multiple ultimately start to come down as perhaps some of the people who would have signed up try to write it themselves?
Jenny Johnson
She added that those would be the smaller companies, which were probably not generating much of the vendors' revenue in the first place.
5. Capital Light No More
The anchor brought in the conversation the program had run with Howard Marks of Oaktree minutes earlier, in which Marks said this is not quite irrational exuberance, and put back his own objection: buying the promise of future profitability is a bet on something unknown. Is AI, in its current form, going to be a profitable technology?
Johnson called that the real question and started with her own accounts.
Her firm's margin has not been rerated by using AI
I actually think for companies like Franklin Templeton, honestly, we haven't been suddenly rerated in our margin, because we're now leveraging AI.
Jenny Johnson
The larger issue she raised is a change in the shape of the AI businesses themselves, and where the cost of that change currently sits.
The AI companies stopped being capital-light, and the spending is not yet in the accounts
these big AI companies who were capital light businesses are suddenly capital intensive businesses as they build these data centers and other things, and a lot of that expense hasn't gone through the p and l.
Jenny Johnson
She then connected it back to her own governance argument: the revenue side is exposed to exactly the discipline she has been recommending.
If customers route more queries to free models, that hits the vendors
as people get smarter about that governance layer on where to direct their queries and start to send more to the free AI models, that could impact.
Jenny Johnson
The anchor added cheaper competing models from China as a second pressure on the same line, and Johnson agreed.
6. Warsh Had To Raise
Asked whether she is satisfied with what she has heard from Kevin Warsh a few months into the job, and what the previous week's meeting told her, Johnson answered on credibility rather than on the rate path.
Nothing about the hawkishness was a surprise
he was always kinda hawkish. I think people wondered how much the president might put pressure on him.
Jenny Johnson
Her reading is that the jobs numbers and the inflation numbers left the Fed no choice if it wanted to be believed, and that the market's reaction was about the institution rather than the decision.
The rise was the price of credibility, and the market bought the independence
To be credible, the Fed needed to raise rates. He's one voice there. And so I think the market's been happy with his independence, his demonstration of the independence.
Jenny Johnson
7. Resilience Is The Word
The last question was a vibe check: given the geopolitical uncertainty and the uncertainty over AI, how is she feeling as an investor?
One word for 2026
if you look at 2026, probably the best word to describe it is resilience
Jenny Johnson
The list of things the economy has absorbed without breaking
corporate earnings are very strong. The consumer continues to be very strong. The job market is strong. Here you have two wars. The oil price is really, really high. We've been able to absorb all those things. Productivity is at an all time high and, again, kinda just starting.
Jenny Johnson
Bonus Insights
The anchor's running joke set the segment's tone
The interview opened with the anchor proposing to take a sip of coffee every time either of them said "AI", and later comparing the new style of programming — specifying what a system must not do — to dating. Johnson played along and kept answering.
She was following Howard Marks on the same set
The program had interviewed Howard Marks of Oaktree minutes before, and the anchor carried Marks's "not quite this irrational exuberance" framing into this conversation as a challenge rather than as agreement. Johnson did not take a position on Marks's view; she answered the profitability question underneath it.
The mainframe statistic belongs to the show, not to the guest
The claim that probably half of financial services firms still run a mainframe in their back office was the anchor's, offered in support of Johnson's installed-base argument. She endorsed it — "they might not admit it" — rather than sourcing it.
Johnson's bottom line is that the AI question facing a chief executive is not whether to slow down but whether anyone is watching what their own agents cost: the one tool she can point to has produced 23% more client meetings and an 11% lift in sales, everything else in the program is still unpaid, and the firms selling the models have moved onto balance sheets heavy enough that the payoff for them is no more settled than it is for her.
Products, Companies & Tools Mentioned
Franklin Templeton (Her firm. Its internal "intelligence hub" directs the sales force at the right clients, producing "23% more clients" and an "11% uplift in sales"; she says the rest of the AI program "hasn't paid off yet" and has not rerated the firm's margin)
Salesforce (The CRM her son-in-law's cleaning company did not buy, having written its own — her example of programming reaching people who previously had to purchase software)
Oaktree Capital Management (Howard Marks had been interviewed by the same program minutes earlier; the anchor carried his "not quite this irrational exuberance" line into this segment as the challenge Johnson was asked to answer)
The Federal Reserve (She says the jobs and inflation numbers meant it had to raise rates to stay credible, and that the market has been happy with Kevin Warsh's demonstration of independence)
Listen to the full episode
🔴 YouTube
Watch the full episode:
More Jenny Johnson interviews and podcast appearances
If this was worth your time, send it to someone who has to have a view on this.
Get the latest market chatter as it happens: