Many pundits suggest that artificial intelligence solutions will soon occupy the same place in business technology as Software-as-a-Service (SaaS) applications do today -- but most fail to note that the process of buying AI is very different from buying SaaS.
Buying AI differs from purchasing SaaS in three key areas: pricing, piloting, and evaluating the solution. If you try to use your SaaS playbook to acquire an AI system, you'll likely buy the wrong product or, worse, pass on an AI solution that you really need.
How AI Pricing is Different from SaaS
SaaS solutions are typically sold on a per-user basis, which allows you to scale up or down based on employee headcount (and avoid the headaches of on-premise software site licenses). You can even adjust your user count month to month, though many SaaS solutions will heavily discount your costs if you commit to annual usage periods.
In contrast, many AI solutions are priced by transaction or completed computation. Modern artificial intelligence software is adept are three broad categories of tasks: prediction, classification, and automation. AI solutions are generally priced by the number of predictions, classifications, or automated functions they perform. The more work AI does, the more you pay for AI.
It's easy to know "how much" SaaS you need, as it scales with headcount (even if you fudge your costs by having employees share logins to your SaaS applications). It may take a bit of experimentation to figure out how much AI is enough for you needs, and success may require more AI than you expect.
The good news is that, thanks to automation, AI is better disposed to help you reduce or repurpose headcount, while SaaS is almost always incentivized to do the opposite (more headcount means more SaaS revenue). AI will likely save you more money in the long run.
How AI Piloting is Different from SaaS
When test-driving a SaaS solution, you typically deploy it to just a few users for a short period of time (usually a 14-, 30- or 90-day trial). SaaS often requires very little in the way of implementation and configuration, and any required data uploads are performed with standard, readily available formats and file types. Moving email, calendar, and CRM data is a known process.
AI solutions usually require specially annotated training data to ramp up, and they need access to as many ongoing sources of data as possible to accelerate their data flywheel. Limiting an AI test to just a few users slows down the solution's ability to adapt to the needs of your business. The more people using artificial intelligence, the faster it learns. Traditional SaaS pilot guidelines, which emphasize small test groups, are actually counterproductive to piloting AI.
How AI Evaluation is Different from SaaS
SaaS solutions, built around freemium models or free trials, are designed to create "wow moments" as soon as possible. The value proposition of a traditional SaaS solution is often obvious -- even visceral -- within hours or days of deploying the solution.
AI doesn't usually have wow moments in the first few days, weeks, or even months of usage. Until an AI system ingests enough training data and is exposed to enough use cases, it likely won't perform any better than conventional software. It's only after AI has been in operation, and its learning curve improves exponentially, that its value becomes obvious.
You simply have to be more patient with AI than you are with SaaS -- but that patience is usually rewarded.
Get Started Now
If AI takes more time, more usage, and more investment to prove out than SaaS solutions, then the sooner you start test-driving AI, the sooner you'll determine whether AI has value for your organization.
If you'd like to learn what artificial intelligence can offer your customer support team, Talla is building the virtual support assistant you should try. To get your AI journey started -- and evaluate AI for support as soon as possible -- contact Talla today.