When I first wrote about practical AI in 2024, the debate centered around whether or not to adopt AI. That question is settled. Our Pulse of Work research, which surveyed 2,500 employees and IT leaders across 10 countries, found that 98% of IT leaders say their companies are now using AI. The harder question is what comes next: how do you make AI an asset, not a liability? Our research shows 70% of employees are using it improperly, and 91% of IT leaders worry AI could make a critical mistake for their business. That’s a substantial risk.
At GoTo, we’ve been building against this problem since before it became a headline. That’s why we developed our five pillars of practical AI – to act as the principles behind every AI decision we make, in our products and inside our own company. We don’t just recommend this philosophy. We run on it.
Here's how we are thinking about each pillar at GoTo, and what they mean for us today:
Pillar 1: AI Must Be Business Impact-Driven
The most expensive AI mistake I see companies make is treating AI as a technology to explore rather than a solution to a specific business problem. In fact, 21% of IT leaders admit their company is adopting AI simply because they think they should. That’s not a strategy, it’s an expensive way to generate activity without outcomes.
Small and mid-sized businesses (SMBs) get this right more often than enterprises. By necessity, they are pragmatic. They want speed and a clear return on investment — that means AI that grows their customer base, serves customers better, or does more with existing people. They don’t have the budget for pilots that go nowhere. If something doesn’t deliver, they move on. That discipline is not a limitation, but a practical advantage. It is a lesson enterprises should take seriously.
The right first question to ask is the same for businesses at any size: what specific problem am I solving, and how will I measure success? For instance, a dealership missing service calls is losing revenue. Or a healthcare practice burying front-desk staff in phone triage is losing time that should go to patient care. These problems are solvable, but only if you start with the challenge, not the technology.
Pillar 2: AI Must Be Ambient
Ambient AI doesn’t announce itself. It shows up at the right moment in your workflow — already there, already helpful, without requiring you to go find it.
Think of Google Maps while you’re driving. Once it’s running, you don’t need to open a separate application for different functions. The guidance is simply there, woven into the experience. That is ambient AI. As an alternative example, consider a technician using an AI copilot on a support call. The right AI copilot can quickly and easily analyze a device, interpret error messages, and suggest resolution steps without being asked. The AI is invisible. It is simply part of how work happens.
AI should fit into existing workflows without requiring behavior change, complex setup, or a dedicated team. This is where the most reliable ROI lives. From there, automation and eventually agentic AI follow naturally, and that is where ROI compounds. AI doesn’t just save time on existing tasks; it redesigns how work gets done entirely. But every step has to be earned. Ambient AI that genuinely works is the foundation.
Pillar 3: AI Must Amplify Human Judgment
AI is taking on more autonomous work. That’s the direction the industry is moving, and it’s the right one, so long as the human-AI relationship is designed to earn that autonomy responsibly.
Trust works the same way regardless of whether a person is working with another human or an AI system: through repeated experience, a visible track record, and the ability to course-correct when something goes wrong. At GoTo, AI starts by working alongside humans — analyzing, flagging, and recommending actions — while the human stays in control. As positive performance is demonstrated, scope expands. No one hands over control on day one. Autonomy is earned, incrementally, through evidence.
Our Virtual IT Technician in LogMeIn Resolve, for instance, follows exactly this arc. In an active support session, AI works alongside the human technician to surface insights, catch things missed under pressure, and recommend resolution steps. Once a solution is proven across enough cases, that technician can confidently let AI handle those issues automatically in the future. Our AI Receptionist in GoTo Connect works the same way: routine calls are handled, and human judgment is brought in the moment it’s needed. The human is always reachable, and the AI earns more responsibility over time.
This trust-to-adoption relationship unlocks the real potential of agentic AI. But it only holds if what’s underneath is observable, controllable, and secure. That is what the next pillar is about.
Pillar 4: AI Must Be Secured and Governed to Be Trusted
For most organizations, AI is missing an important governance foundation. Fewer than 44% have a formal AI policy in place. 83% of employees worry they could be blamed or fired for an AI mistake. 14% say they reported an error and were told to stay quiet. This governance gap is not closing on its own — and in many organizations, it is being actively hidden.
At GoTo, trusted AI means three things. First, transparency: customers always know when AI is involved, what it is doing, and why. It’s not a disclosure checkbox, it’s the confidence to set meaningful boundaries. Second, control: customers decide how far AI goes. Are you comfortable with an AI receptionist handling all calls? Or do you want to compare its performance before expanding its scope? Both are valid. Our job is to make that choice easy, visible, and reversible.
Third — and most critical in 2026 — is agentic guardrails. These are explicit limits on what AI can do without human approval. AI agents chain actions across multiple steps, and a wrong decision compounds quickly. LLMs hallucinate. Agents go off-script. The question every organization needs to answer is not just “what can our AI do?” but “what are we comfortable letting it do without asking us first?” We answer that question at the architecture level — not just by policy, but by how the systems are built. And we’ve found that getting governance right early doesn’t slow you down. It’s what makes it safe to move fast.
Businesses trust us with their data and to let AI act on their behalf. We take that seriously in how we test, monitor, and constrain AI behavior, and in how transparently we handle customer data. Security and governance are not compliance obligations. They are the foundation on which every other pillar stands.
Pillar 5: AI Must Be Tailored to Your Business
Two groups are always involved when implementing AI: the technical experts who build or deploy the solution, and the end-users who work with it every day. Tailored means getting both right — and they need very different things.
For technical experts, complexity scales with the organization. Small teams need turnkey solutions that work immediately, no configuration required. Mid-sized organizations need customization. And large enterprises with experienced IT and engineering teams need full agent orchestration and complex workflow integration. The goal is to match the implementation to the real technical capability in the room. Don’t oversell complexity, but don’t undersell what experienced teams can handle either.
For end-users, none of that complexity should show. Whether they are in a five-person practice or a 50,000-person enterprise, all end-users want the same things: AI that is simple, predictable, easy to use, and seamlessly integrated. AI succeeds or fails based on whether people will work with it. Technical sophistication belongs in the infrastructure. The surface must be designed for adoption.
The same logic applies to the larger industries adopting AI. Our Automotive and Healthcare SKUs are not generic AI with a new label. They are purpose-built for the specific workflows, integrations, and regulatory requirements of those industries. AI Service Scheduling fills service bays and syncs with dealer systems. AI Healthcare Appointment Scheduling connects directly to EHR platforms. They are tailored to your industry, ready out of the box.
The Outcome: Orchestrated Intelligence
When these five principles work together, AI stops being a collection of features and becomes the way your organization operates. We call that Orchestrated Intelligence: AI deployed with intention, matched to workflows, enabling people effectively, and governed without scaling risk.
We don’t just build these principles into our products. We run on them. GoTo’s engineering and operations teams work inside the same framework — the same five pillars, applied to how we build, manage infrastructure, and govern AI internally. You cannot credibly sell an organizational vision you haven’t lived. This is ours.
The advantage isn’t access to AI. It’s knowing how to put it to work.