The best AI advice at RetailNOW 2026 didn’t involve a clever prompt trick. It didn’t require a data science team, an enterprise platform, or a plan hammered out in a boardroom, either. It required something simpler: Pick one workflow employees already complain about, decide what data is safe to use, put one person in charge, measure performance before and after, and expand only if the results earn it.
That was the throughline of the July 27 RetailNOW 2026 session “What VAR & ISV Teams Should Learn Next About AI.” The panel was moderated by Audrey Hogan, vice president of marketing at LINGA, and included Tigran Mikayelyan, president and CEO of CoCard; Travis Hare, vice president of customer success at LINGA; Ashton Radican, vice president of sales for LANDI Global’s western region; and Blake Gillum, president of DCR.
Using AI isn’t the same as being AI-ready
The panel framed AI and AI-readiness as the difference between knowing how to drive and completing professional driver training. Anyone can open ChatGPT or Claude and get useful output. That’s using AI at the consumer level. An AI-ready business has approved tools, data rules, defined accountability and a structured way to roll all of that out. One is a skill. The other is an operating discipline.
Radican put a finer point on what separates a good AI user from an average one: “Knowing what to ask, knowing what to check, and the one people forget is knowing when just to do it yourself.” That’s a more useful bar than counting how many prompting tricks someone knows. It puts the responsibility back where it belongs – on the person supplying context and reviewing the output, not on the tool.
Gillum described building a personal “second brain” (an AI assistant paired with a notes tool like Obsidian) that he says returns hours to his day by organizing documents, tracking customer history and surfacing what he needs before a meeting. The catch: it only works inside guardrails. Approved tools, approved data. The moment customer information enters an unapproved system, the personal-productivity win turns into a liability.
Start with one problem, not a platform
Hare argued that how you prep the AI matters more than how cleverly you prompt it. Feed it clear context and a real objective, he said, and you can talk to it almost conversationally and still get strong, reliable output. Skip that step and the same tool will confidently produce nothing useful. His broader warning: adding AI to a broken or undefined process doesn’t fix it. It just automates the confusion faster.
Two other myths came up repeatedly. First, that AI implementation requires a data science team, a big budget and expert staff. The panel’s counter was blunt: you don’t need to be perfect, you need to take a first step. Second, that dumping unstructured information into an AI chat and expecting a coherent answer back is a strategy. It isn’t. Garbage in, as Hare put it, is still garbage out.
The panel’s prescribed starting point: one small task, one person, a workflow the team already complains about. Simple enough to start on a Monday and show impact by Friday. That’s not a slogan, it’s a way to build evidence before committing budget to something bigger.
For sales teams, that might mean prospect research, call summaries, CRM updates or follow-up cadence reminders, with the salesperson still owning the relationship and the decision. For support, it’s ticket summaries, categorization and turning resolved tickets into a searchable knowledge base, with AI assisting agents rather than answering customers directly. Mikayelyan offered the session’s most eye-opening data point: AI has helped CoCard compress its product-development cycle from four years to six months.
In MSP terms, the strongest early candidates look similar: PSA and ticket documentation, knowledge capture from resolved tickets, quarterly business review prep, proposal drafting, and account research before a client call. These are repetitive, measurable, and low-risk – someone reviews the output before it ever reaches a client.
Put governance before scale
Mikayelyan laid out a four-part governance model that gives channel companies a practical starting point:
- Approved tools: Tell employees which AI products they’re authorized to use and how access is managed.
- Approved data: Define what information can and can’t be entered into AI tools. Treat it the way you already treat confidentiality and data-security policy.
- Accountability: Identify who remains responsible for customer-facing, financial and operational decisions influenced by AI.
- Tool ownership: Assign someone to review use cases, manage permissions, and handle incidents and updates for each approved tool.
None of this needs to be a 50-page legal document. The panel favored a one-page policy that’s socialized (i.e., talked through with employees) rather than just filed away, and revisited quarterly. Before writing it, consider an anonymous survey to find out which teams are already using AI, which tools, and what they wish they could do with it. Most organizations have more shadow AI use than leadership realizes, and a survey turns that discovery into a planning input instead of a disciplinary conversation.
For MSPs specifically, the stakes are higher. Tickets and documentation routinely contain credentials, configurations, PII and data spanning multiple clients. That means understanding how each AI provider retains and uses submitted data, keeping client data properly separated, and requiring human sign-off before any AI output triggers a customer-facing or operational action.
Measure time saved without losing sight of quality
The panel’s simplest metric: time saved. Measure how long a task takes today, run the pilot, measure it again. It’s provable, and it’s the number that justifies the next pilot. Beyond that, there’s no need to invent new KPIs. Indeed, time in queue, ticket-resolution time, sales-cycle length and proposal turnaround still apply.
Pair time savings with a quality check, though. Ticket reopen rates, first-contact resolution, documentation accuracy or proposal win rate all tell you whether faster work is actually better work or just faster mistakes.
Make sure AI agents can actually find you
The panel also touched on answer engine optimization for making sure AI agents, not just search engines, can find, understand and accurately represent your business when a prospect asks. As buyers increasingly query ChatGPT or Claude instead of Google, that’s a real gap for VARs, ISVs and MSPs whose sites were built for the old rules.
A quick first pass: ask an AI tool to review your website from the perspective of an agent trying to answer a prospect’s question, and to flag what’s missing or unclear. Treat that as a starting audit, not a full strategy. You’ll still need clear service pages, credible evidence, and content that directly answers the questions your prospects are actually asking.
A 90-day AI roadmap for solutions providers
Days 1–30 – Identify and control: Survey current AI use, pick one recurring workflow, document its current performance, and define the approved tool, permitted data, owner and human-review step.
Days 31–60 – Pilot and measure: Test with one person or a small team. Track before-and-after time alongside at least one quality metric. Document errors, security concerns and employee feedback.
Days 61–90 – Decide and scale: Determine whether results justify expansion. Update the one-page policy, train the next group of users, and schedule a quarterly review of tools, access and outcomes.
Once a solutions provider has proven this internally, it becomes a client-facing service: AI-readiness assessments, acceptable-use policy workshops, data-readiness reviews, limited workflow pilots. The value isn’t in reselling another tool. It’s in helping a client pick the right problem, protect their data, and prove an outcome.
AI readiness was never really a technology purchase. It’s a management discipline built on useful questions, clean data, clear accountability, and a number you can actually point to.
