Salesforce showcased its new Agentforce platform in San Francisco last week, promising businesses a new kind of autonomous business assistant.
The company says AI agents could revolutionise the way some of the world’s biggest enterprises operate.
SmartCompany investigated how smaller businesses can plug into the new AI system.
Here’s what we know so far, based on conversations with senior Salesforce personnel and businesses already toying around with Agentforce technology.
Probably not — and Salesforce doesn’t claim they will.
If your business isn’t working now, the addition of AI agents is unlikely to fix it.
“As we’re talking about AI, as we’re talking about agents, I think we have to step back and always realise that there’s no panacea here,” says Ryan Gavin, chief marketing officer at Slack.
“If you have an inefficient way of working and you’re just bringing in AI agents, you’re just going to magnify the inefficiencies… You have to step back, and you have to have a new way of working in this age of AI agency, and be really thoughtful and deliberate about that.”
Probably not, at least for now.
Salesforce CEO Benioff envisions AI agents augmenting and supporting human workers, allowing them to focus on high-value tasks like customer or patient care instead of administrative grunt work.
In one example, Benioff points to the use of agents in a Gucci call centre handling repair requests.
The agent provided staff of the luxury brand advice on upselling opportunities, increasing sales and turning an “expense centre” into a “revenue centre”, Benioff claims.
In that scenario, where agents empower workers to be markedly more efficient, reducing headcount might not make sense.
For small businesses, the likelihood of AI agents outright replacing staff is even slimmer.
Salesforce says agents can make annoying or repetitive tasks a little faster, and a little simpler.
But the remaining tasks won’t disappear, and someone in your business still needs to do them.
Secondly, there’s the per-user, per-month payment model inherent to many of Salesforce’s entry-level plans.
Naturally, Salesforce wants more small business employees using its systems under that model, not fewer.
Customers “not staying with us is our kryptonite; it’s the worst thing that could possibly happen,” says Kevin Doyle, regional vice president of data and AI for the APAC region.
A successful startup using Salesforce tools might “hire a second person, so now it’s two users per month, and then three per month, and then they’ll want to use more Salesforce products,” he says.
“So customer success is so deeply embedded in how we need to be successful as a company, and it starts with the one-person startup.”
Mass layoffs because of agentic AI seem extremely unlikely in the short term, but Salesforce envisions some changes to seasonal workforces.
Using the example of a major textbook publisher, Benioff says the company could feasibly “surge” its workforce in the back-to-school rush by implementing AI agents.
In that scenario, the use of agents might reduce seasonal hiring, but might not carve into permanent staffing numbers.
Doyle says the technology may also influence future hiring decisions, without directly affecting those already hard at work in a small business.
“Instead of me hiring a large marketing team, and a large customer service team, and a large sales team, I can now have agents augmenting my team to allow a fraction of the employee size to the same amount of job,” he says.
Driving the point home, Doyle describes agents as a way to do more with the same number of staff, as opposed to doing the same amount with fewer staff.
In small businesses, “you’ll never be able to outspend the competition, so you have to outserve them,” he says.
“So how can you have a leaner, meaner force, using AI, using agents to augment your work, to drive better outcomes?”
Yes.
Jayesh Govindarajan, senior vice president of AI and machine learning at Salesforce, tells SmartCompany businesses can build an agent without feeding it pre-existing company data.
But to do so, a business must carefully provide a role and parameters to its agent.
Govindarajan uses the example of a new startup with no customers, but a strong understanding of its target market persona — a stand-in for who its real customers could be.
The startup can feed its agent a careful description of the marketing persona, and “four or five really high-value problems to try and solve”, like creating a successful campaign.
To allow the agent to act on those jobs, the startup can link it to an LLM capable of creating marketing communications, and systems capable of retaining and labelling inbound data.
“Now you’re starting to use your expertise in understanding the problem, and break it down into data, and into a set of actions that you get done,” says Govindarajan.
With those settings in place, the startup can put the agent-assisted marketing plan in place, assess its performance in the real world, and tweak the model accordingly.
“That’s really all you need to get started,” says Govindarajan.
He describes a feedback loop: more real-world data means more material for the startup to assess and feed back into the agent’s campaign-building model.
“You build it, you launch it, and then you listen to your customers as a startup to then add new features,” says Govindarajan.
Agentforce will become generally available for Sales and Service users on Enterprise plans and above from October 25, 2024.
Businesses in Sydney and Melbourne can get a hands-on test from October.
The author travelled to Dreamforce in San Francisco as a guest of Salesforce.
Never miss a story: sign up to SmartCompany’s free daily newsletter and find our best stories on LinkedIn.
Senior Business Journalist
Focused on the small business sector, David’s work covers the political, regulatory, and economic issues facing Australian entrepreneurs. Prior to joining SmartCompany, he was a reporter for Business Insider Australia. You can follow him on LinkedIn.
