It wasn’t too long ago that the AI conversation centered around adoption, but that conversation has changed. Today, most Salesforce customers accept that AI will have its role in their future, but the question now is what that future actually looks like.
Some are considering building their own AI stacks with tools like Claude and OpenAI, while others are buying AI-native products to solve problems specific to their business. Many are also putting their faith in Salesforce’s own roadmap through Agentforce and Data 360. None of these paths are necessarily wrong, but they do come with their own trade-offs around cost, speed, and flexibility.
Most businesses seemingly have three options, and it feels like less of a technology decision and more of a business bet. The companies that get it right could gain a competitive advantage, but those that don’t risk spending huge amounts on AI without seeing meaningful returns.
What Are Salesforce Customers Actually Doing?
At the moment, Salesforce customers are trying to determine the best way to invest in AI. A significant hurdle that comes with this is actually finding business value and selling the prospect of AI to stakeholders.
Mike Piehl, CEO of Platinum River Innovations, has watched that journey play out for the last couple of years.
He told SF Ben that many larger companies initially went all in on building their own AI tools, purchasing large language models to run behind their own firewall. While the approach offered greater control, it also came with high costs and disappointing returns.
He said: “At that point, all of our larger clients were purchasing LLM models to train and run within their own firewall. That approach was expensive, both with LLM costs and Data Scientist salaries, and yielded few results.”
The phase that followed this wasn’t much better. Rather than building from scratch, many businesses turned to what Mike described as “popcorn” AI tools – quick proof-of-concept projects designed to solve isolated problems.
These were faster and cheaper to implement, but often failed to make an impact. “Those two phases of AI adoption resulted in AI burnout in many of our clients,” Mike said.
From this, Mike believes that customers are taking a more measured approach. Instead of chasing the latest model or AI product, they’re thinking more carefully about where AI actually fits and whether their future lies in building, buying, or betting on Salesforce’s AI roadmap.
Bet 1: Build
Despite some of the aforementioned risks, building your own AI stack still makes the most sense for some businesses.
Using models like Claude or OpenAI gives businesses greater flexibility, allowing them to tailor AI around their processes rather than waiting for a vendor’s roadmap. For companies with more complex requirements or systems extending beyond just Salesforce, having that control gives them a big advantage.
But Mike argued that many businesses underestimate what’s required to get there. In his experience, successful AI projects often depend less on the model itself and more on the quality of the underlying business and data.
“Many AI providers pitch AI solutions in weeks,” Mike said. “However, to achieve AI readiness, there is a very large effort to remediate 10+ years of technical debt. So a $10K, two-week AI rollout actually takes $1-2M over 12-18 months for a 500-2,000 employee company with legacy software systems.”
In essence, you may have long-term flexibility if you’re looking to make that investment. But cost, technical debt, and ongoing maintenance mean it simply isn’t the quickest or safest route to gain value from AI.
Bet 2: Buy
Not every business will have the time or resources to build its own AI stack. Instead, the next best bet would be to look toward AI-native vendors that solve specific problems – whether that’s customer service, sales prospecting, document processing, and so on.
The appeal of buying is pretty obvious. Specialist tools can be deployed much faster than a large transformation project, allowing businesses to prove value before making any larger investments.
However, Mike believes organizations often make the mistake of treating AI as the starting point rather than the solution.
“AI, like all tools, starts with why. However, unlike recent tools, AI has the ability to fundamentally restructure businesses.”
He argues that businesses should first understand where AI can create meaningful value, whether that’s improving customer experience, reducing costs, or creating new revenue opportunities, before deciding which technology best supports those goals.
Justin Piehowski, Lead Salesforce Advisor at The Hatch Group, also believes that too many organizations are still evaluating AI through the lens of software procurement rather than business strategy.
“Instead of asking, ‘How can AI help us create new value or compete differently?’ they’re asking, ‘How can we bolt AI onto the business we already have?’ That’s a much narrower view of the opportunity.”
Bet 3: Trusting Salesforce
For the majority of Salesforce customers, the third option is arguably the most straightforward, which is to build around Salesforce’s current AI vision.
This hasn’t been without its challenges, of course. Since Agentforce launched, questions around pricing, implementation complexity, and return on investment have left a portion of Salesforce customers cautious and apprehensive. Like much of the AI market in general, a lot of people are still trying to separate genuine business value from the hype.
Yet despite the concerns, Salesforce still holds the advantage as they hold on to the data. Customer records, sales pipelines, service history, permissions, automations, and increasingly unstructured business content already live within the Salesforce ecosystem. Rather than moving that data elsewhere or stitching together multiple AI tools, many organizations see value in bringing AI to where their business already operates.
Justin says that familiarity is already influencing customer decisions, saying: “Most customers in the Salesforce ecosystem aren’t really making a pure build-versus-buy decision. They’re intrigued by AI and experimenting across the board, but I see many organizations gravitating toward platforms they already trust, including Salesforce.”
That doesn’t necessarily mean Salesforce will emerge as the long-term winner. AI is evolving too quickly for anyone to say that with certainty. But for organizations looking to balance governance, security, and existing technology investments, betting on the platform they already know may feel like the least risky move, even if it isn’t the most adventurous.
Final Thoughts
AI isn’t going anywhere, and for Salesforce customers, the time for simply watching from the sidelines is coming to an end. As I covered earlier this year, organizations that rush into AI without a clear strategy risk wasting significant amounts of money. But doing nothing is quickly becoming just as risky.
Ultimately, every customer is placing a bet. Building offers flexibility, buying offers speed, and Salesforce offers familiarity, governance, and access to the data many businesses already rely on.
For me, Salesforce still holds the advantage. The question now is whether the mothership can continue giving customers enough compelling AI use cases to justify that bet. It has a significant head start, but standing still isn’t an option in a market moving this quickly.
