(This column originally appeared in Forbes)
Run a business? Want to learn what enterprise AI is coming that will impact your business?
Then do what I do and follow the activities of Emergence Capital. They’re a venture capital firm focused on early-stage enterprise software companies and have been behind some of the most disruptive platforms used by your business and mine like Salesforce, Box, Bill and Zoom.
Of course most of the conversation around technology nowadays is about AI and how it will be impacting businesses. That’s Joe Floyd’s job. A general partner at Emergence, Floyd has spent over fifteen years advising and investing in “rocket ship” startups like SalesLoft, project44, LogDNA, and other similar companies.
So how will AI impact your business this year and in the near future? Here are some of his thoughts.
Enterprise AI Voice is ready for prime time.
According to Floyd, the advent of reading models is getting extremely good.
“Prior, you really couldn’t have a full fledge voice AI agent that sounded human,” he said. “And now for the first time, you can, and it’s a combination of sub-500 millisecond latency. These really human sounding voices have the ability to reason, understand where you are in the conversation, and then actually have the appropriate response.”
Floyd says that AI voice agents are now outperforming humans.
“They stick to the call script repeatedly and they never get tired,” he said. “They never have a bad moment. They’re doing it at a rate that is higher than the average success rate for humans. They just execute.
All of this is happening very fast, according to Floyd, and the winning companies will be the ones that are laser focused on strategies like: “I’m going to be the best voice agent for mortgage servicing and mortgage sales” or “I’m going to be the best at healthcare scheduling.”
“There are real cost savings for businesses using AI voice agents,” he says. “The ROI is just undeniable.”
Agentic workflows will explode.
Floyd acknowledges that AI agents are — obviously — a big area right now and he’s looking closely at companies that are building proprietary models that are improving workflows faster using agents.
“We’re seeing a lot of AI startups and companies that are taking over a manual process with agents, like document ingestion or communicating with customers to answer questions or perform compliance,” he said.
Will these agents ultimately replace humans in the workplace?” Floyd says it depends.
“For example, AI agents will help accountants do all the boring stuff that they don’t want to do, so that they can actually have the interesting strategic conversations that they want to have — and also serve multiple customers with the same staff,” he said. “Smart companies are providing tools to create smaller large language models that are hyper-tuned to do one thing and a larger model can choose one of maybe forty different agents to get a specific job done.”
Floyd is also seeing a proliferation of voice agents being used for sales and customer support.
“These are things that both enterprise and small business will definitely adopt,” he said.
A more efficient use of infrastructure is coming.
There’s still seemingly an unlimited demand for data centers and server space that can’t be immediately met. But Floyd doesn’t think the need to build bigger and more data centers will continue in the long term. Some startups are focusing on making the data better.
“At some point you’re actually not going to need to build bigger and bigger large language models,” he said. “What we’re seeing is the post-training of large language models starting to take a much bigger role than pre-training. Models are staying the same size, but are being pruned, not tuned, to make them better and there’s more of a premium placed on the quality of data, rather than size of the infrastructure.”
AI will be taking over the desktop.
Tech companies from Google to Anthropic are already introducing AI applications that can take over a device and perform web browsing activities with minimal human involvement. Floyd says this will continue to evolve over the next few years.
“You can imagine a world where there’s a website where humans can read and there’s also a website version that is machine optimized and readable and then the software just goes to town without human involvement,” Floyd said. “As part of the operating system, an AI application is able to capture anything on the screen, read it, process it, and have it accessible to a large language model. It will have its own small model built in, which means that my data’s never going out to any public servers unless I allow it to. So I can say, ‘hey, this requires a level of, of, of understanding that isn’t capable of the small model that’s running on my CPU. Go ahead and, and send this information out. Get me a better answer.’”
Quick building enterprise AI apps are starting to come of age.
Just a few years ago new applications from web building platforms allowed small businesses to create their own websites for the first time ever and minimize the need to hire an expensive web developer. But these were still complicated and the average owner was still not really cable of using them. Floyd’s excited by new agents that can enable those same business owners to do this by talking in natural language and building websites on the fly.
“This is only the start,” he said. “Web applications now have their own databases that can enable small business owners to build their own applications — like a CRM or project management system — in the same way.”
Using one of these platforms Floyd — who says he hasn’t written a line of code in “20 years” was able to build his own CRM application “in under 30 minutes!”
Old school vendors are partnering with enterprise AI startups.
Floyd is not confident in the ability of legacy software vendors to build their own AI functionality.
“I think they will have absolute garbage and I don’t think any of them will do anything interesting,” he said. “It is already being proven out that the startups are simply producing better, more highly tuned solutions.”
So what are older vendors doing? According to Floyd, the smarter ones are partnering.
“AI startups need data and distribution which the old companies have,” Floyd says. “And you’re seeing more of these partnerships than I think you ever saw before. I don’t think the legacy vendors will be able to build good AI solutions themselves. Some will be able to buy or partner their way into decent solutions.”
A final warning for business owners.
Floyd warns that if you’re running a company and you’re not adopting AI, your company just won’t be around in five years.
“Even if you are a nuts and bolts company that you think has nothing to do with technology your competitors will adopt it and they’ll out compete you,” he said.
Floyd says that small and mid-sized business owners are the ones that need to pay attention to AI most.
“AI agents are perfect for a lot of manual uses cases — like order entry, inventory management, answering phones, sending invoices, collecting or disbursing payments,” he said. “One company I came across recently is helping physical store owners parse invoices for goods ordered and input that data directly into inventory systems. This is saving owners hours of manual work every week and they are doing it at a fraction of the cost. I think there will be lots of enterprise AI use cases like this for owners and employees of these smaller companies.”
