This is the published version of Forbes’ CFO newsletter, which offers the latest news for chief finance officers and other leaders focused on the budget. Sign up here to get it delivered to your inbox every Tuesday.

As 2024 moves into its second half, many CFOs have the same top strategic priority: Using AI and other automation techniques to do tasks previously performed by humans. The Q2 edition of the CFO Survey, done by the Federal Reserve Banks of Richmond and Atlanta and Duke University’s Fuqua School of Business, found that nearly six in 10 companies have already implemented software, equipment or technology to perform these tasks, while 53.6% seek to add to their tech stack to automate more employee tasks in the next year.

However, it’s important to note that most financial departments aren’t looking to replace those humans, for now. More than 87% that implemented AI in the last 12 months did it to enhance their business processes—just about the same proportion that plan to use AI for this function in the next 12 months. Increasing output quality is also key—57.7% added AI for this reason in the last year, while 58.1% plan to get that outcome in the next year. Less than half wanted to reduce labor costs through AI in the last year, but 54.7% stated that was a goal for the next year.

Daniel Weitz, survey director for the Federal Reserve Bank of Atlanta, writes that the current economic situation might be pushing CFOs more toward AI automation. As inflation grew in 2021 and 2022, input costs and employee wages also increased, which in turn led to higher unit costs. The pressures of the recent past have moderated, but pricing pressures are still pretty high. Survey results show that companies that decided to automate tasks last year also experienced faster price growth then. And those same companies are anticipating slower price growth this year, perhaps because automation—and greater accuracy that comes with machine-based calculations—means the company spends less.

“Time will tell whether the bifurcation of price growth expectations by automation status is borne out, and whether automating firms return to ‘normal’ price growth more quickly than their non-automating peers,” Weitz wrote.

Most business enterprises assume that AI will help their bottom lines, but figuring out how much—and how long that will take—can be a challenge. I talked to David Obrand, CEO of AI-powered sales engagement platform Salesloft, about the aspects you should look at when making that calculation. An excerpt from our conversation is later in this newsletter.


The last week has been a series of huge increases and moderate decreases on the stock market, mostly thanks to Nvidia. The S&P 500 hit 5,500 for the first time ever last week, and many tech stocks surged as AI chip maker Nvidia briefly became the world’s most valuable company last Wednesday. It ceded that crown back to Microsoft on Thursday and fell to No. 3 behind Apple by Friday. Following last week’s meteoric rise, Nvidia’s stock has been quickly losing steam, dipping below $120 a share this week.

Nvidia is the latest beneficiary of AI mania on the stock market, as excitement over the potential AI technology has helped many big tech companies reach new highs. Apple—which announced soon-to-come AI integrations for its phones, tablets and computers—and Microsoft—which received a favorable report from analysts—started the rally that pushed Nvidia into the stratosphere last week. But Nvidia also had its own big news last week: a partnership with Hewlett Packard Enterprise for a portfolio of jointly developed AI solutions and integrations to help enterprises quickly adopt generative AI.

While other companies are making AI platforms, Nvidia creates the chips needed to run those platforms and data. CNBC has estimated Nvidia controls between 70% and 95% of the AI chip market. Bloomberg supply chain data estimates that Microsoft actually makes up 15% of Nvidia’s revenue. Sequoia Capital estimated in March that companies have spent $50 billion on Nvidia chips to train LLMs, the Wall Street Journal reported.

And that doesn’t include Nvidia’s soon-to-come Blackwell platform, which features the most powerful chips ever created. At the time they were announced in March, Nvidia CEO Jensen Huang said the company already had commitments to use the new chips from Amazon Web Services, Dell Technologies, Google, Meta, Microsoft, OpenAI, Oracle, Tesla and xAI. So even if AI platform development hits a slow patch, Nvidia is set to continue its rise in valuation and revenue. Nvidia’s annual stockholder meeting will be held Wednesday, meaning there may be much more announced that can move markets.


The U.S. Supreme Court ruled last week that individuals who own portions of overseas companies can be taxed on those earnings, even if that money has stayed in that business and has not been repatriated. Primarily at issue was whether Congress had the authority to make this law, which was challenged by a couple who owns a stake in an Indian company. The couple argued that the law taxes them on personal property, not actual income.

A ruling in favor of the couple—ordering that they could not be taxed on unrealized income—could have had a huge ripple effect on general taxation in the U.S., likely leading to the federal government missing out on trillions of dollars. Forbes senior writer Kelly Phillips Erb spoke with several voices in the corporate taxation world about the ruling. Many applauded the ruling, if only because it left the current system in place. However, some expressed trepidation that the Supreme Court was interested in hearing a case on the issue in the first place, and felt that further cases that could make changes to taxation may be on future dockets.


The IRS plans to deny tens of thousands of improper high-risk Employee Retention Credit claims made by businesses in the aftermath of the Covid-19 pandemic. Forbes senior writer Kelly Phillips Erb writes this program was intended to help businesses survive the early portion of the pandemic, during which many businesses were forced to temporarily close. Since its inception, businesses have filed 3.6 million claims for this program, which cost more than $232 billion. The IRS says it’s so far investigated 1 million claims worth $86 billion, identifying between 10% and 20% as being high-risk, with “clear signs of being erroneous claims.” An additional 60% to 70% of the claims show an “unacceptable level of risk,” and the agency plans to give those deeper analysis. The IRS says these claims are complex and investigating them takes time, but companies that feel they may have submitted questionable claims can file a withdrawal.


Salesloft CEO David Obrand On How To Figure Out ROI for AI Platforms

When you’re planning to add an AI platform to your business enterprise, it’s difficult to quantify the ROI. I talked to David Obrand, CEO of AI-powered sales engagement company Salesloft, about some of the places that you can start to look to come up with some realistic figures. This conversation has been edited for length, clarity and continuity.

Generative AI platforms come into enterprises to help people do their jobs better. But how do you put numbers on that? If somebody is coming to you and saying, ‘Look, we want to do this, but we need figures,’ how do you do that?

Obrand: We look at the KPIs that are top of mind for any go-to-market organization. That is everything from the cycle time to actually complete a transaction, the size of that transaction, the win rates around that transaction. All of those, that’s what our customers are looking for, especially in more challenging macroeconomic environments that we’re in today, where interest rates are rising, companies are spending less, they’re more guarded about their capital allocation. The bar is much higher right now than it has been in recent years to drive revenue for any type of organization. Everybody right now is looking at: How can we be more effective with fewer resources? And the answer to that is: How can you identify more opportunities being able to ingest digital signals; buyers today are more digitally empowered than they ever have [been] before.

When companies are looking at where they’re going to get their value with AI, is there anything that they tend to forget to look at?

It depends on the use case. In order to train an AI model, you have to ingest a significant amount of data. So you have to be very thoughtful about where that data is coming from, what the data privacy laws are around it. In certain use cases, you have to be mindful of the ethics around the use of AI. You also have to make sure that you are training the users on how to use it properly. I think that is one big area that we are starting to see today.

There are numerous examples of when AI is deployed quickly without users being adequately trained on how to do it. AI, in and of itself, is not tethered to the truth. It’s just trained on [data]. You could apply generative AI to have listened to our conversation, and then write follow-up notes. We actually generate three templates of every call. We allow the sellers to go review those templates, validate, pick and choose pieces they want, and combine it and execute. Imagine if we just generated an email summary, and it was just one version of it, and nobody looked at it. They just sent it off. If the AI had hallucinations, or simply missed core components, you would get an email from me and think, ‘Oh my God, this guy didn’t pay attention to anything I said, and he totally doesn’t get it.’ Then I’m done! So you really have to make sure that you are training the users of the technology in whatever form that it may come in. I think data privacy on one side and user training on the other side, are areas that people in all of their exuberance might shortcut to their detriment.

What kinds of arguments would a CFO expect to hear from someone trying to show that an AI platform is a worthy investment?

In our world, what you would play up would be the ability to generate revenue with far greater consistency and repeatability than you had in the past. As you break that down into silos, you would look at it and say, ‘Look, you can reduce your marketing spend and generate more pipeline than you did previously.’

The second area would be: You could generate more revenue with fewer resources. Now, AI does not replace people, but it does allow people to perform better in their role. If everybody is performing at an elevated rate, I don’t need quite as many of those individuals. You could look at it as revenue per sales employee, for example.

If you’re in a highly competitive market, you could look at it as win rates. Market share. Are we able to garner both a greater market share? Also, because we can ingest first-party data, which is wildly helpful to our customers as they cross-sell and upsell to their customers, you could look at it as: What are my share of wallet improvements?


A new report by Citigroup found that the banking industry is set to be profoundly impacted by generative AI.

54%: Amount of time now spent on banking work that could be automated

$170 billion: How much the study estimates AI could add to banking sector profits by 2028, a 9% increase

‘Revolutionize the banking industry and improve profitability’: What David Griffiths, chief technology officer at Citi, says AI can do


As the CFO’s job has been changing, so has the career path to go from a finance department employee to the top slot. Here are some tips to find your way there from experienced CFOs.

It can be difficult to be productive at work. Here are some tips to increase what you’re able to get done.



Danish pharmaceutical company Novo Nordisk announced this week it is going to build a $4.1 billion manufacturing facility in the U.S. for its weight loss drugs Ozempic and Wegovy. Which state will this facility be in?

A. North Carolina

B. Texas

C. Kansas

D. New York

See if you got the answer right here.

Leave a Reply

Your email address will not be published. Required fields are marked *