New Jersey business leaders gathered virtually on July 29th for a 90 minute NJBIZ panel discussion on artificial intelligence, and the conversation that followed offered a useful corrective to the breathless pace at which most companies encounter the technology. Moderated by BridgeTower Media East Business Division content lead Ben Jacobs, the panel brought together five professionals working across law, accounting, marketing, and human resources, each approaching AI adoption from a slightly different angle but converging on a shared message, that the technology itself is rarely the hard part.

Strategy Before Shopping

Every panelist returned, in one form or another, to the same warning: companies get into trouble not because they pick the wrong AI tool, but because they skip the step of figuring out what problem they are actually trying to solve. Stout described the constant churn of new AI products as genuinely tempting, but argued that the real work lies in first identifying which internal workflows are redundant or wasteful, then evaluating tools against that specific need rather than adopting technology first and hoping a use case follows. Skipping that step, he said, tends to produce a faster version of a bad process rather than a genuinely better one.

Wakiyama made a related point about risk tolerance, noting that appetite for AI adoption varies considerably by industry and by company culture. Highly regulated fields like law, finance, and banking tend to move more cautiously given the legal exposure involved, while other organizations are more willing to experiment and push further out ahead of the curve. Across both types of clients, he said, he consistently advises defining specific use cases before selecting a tool, since skipping that step tends to invite legal and operational risk that could otherwise be avoided entirely.

What Companies Are Actually Using It For

Panelists pointed to a fairly broad set of practical applications already in use across New Jersey businesses, including summarizing lengthy documents, building spreadsheets, conducting an initial screen of job applications, supporting recruitment and background checks, managing social media accounts, and refining marketing demographic targeting. Every panelist stressed the same caveat regardless of the specific use case, that AI generated output needs to be verified for accuracy, sourcing, and completeness before it gets relied upon in a professional setting.

Sailer framed the current moment as fundamentally about training people to use the technology well rather than rushing toward more advanced applications before the basics are understood. She described the underlying goal as freeing employees to use their time more effectively rather than replacing them, framing AI adoption as an elevation of existing roles rather than a threat to them. Douglas offered a related but distinct framing, suggesting that AI’s real value lies in freeing people to focus on the distinctly human parts of their work, and that the businesses that win in the coming years will be the ones that pair strong technology adoption with genuine human differentiation, rather than leaning on one at the expense of the other.

The Free Tool Trap

One of the panel’s more pointed warnings concerned free, publicly available AI platforms like ChatGPT and Copilot. Wakiyama explained that the terms of service governing these tools tend to heavily favor the provider, leaving companies with little practical recourse if something goes wrong. He described fielding calls from clients after an employee had entered confidential business information into a public AI tool, only to learn there was no real way to extract or remove that data once submitted, a scenario that can permanently compromise trade secret protection. His broader advice was straightforward: free tools can be useful in the right context, but every organization needs a clear internal policy spelling out which tools are approved, for which use cases, and under what circumstances information can be shared with them.

Wakiyama also cautioned that the legal and regulatory landscape around AI remains genuinely unsettled, and is likely to stay that way until Congress meaningfully addresses it. In the meantime, he described the environment as a patchwork of overlapping laws, industry specific regulations, and case law, making strong contractual protections with AI vendors and data partners one of the more reliable tools companies currently have to manage their legal exposure.

Building Real Governance, Not Just a Policy Document

On the question of how companies should actually structure AI oversight internally, panelists agreed that governance works best as a cross functional effort rather than something owned by a single department. Wakiyama suggested involving representatives from legal, human resources, finance, marketing, sales, R&D, and IT wherever an organization’s size allows for it, while stressing that even smaller companies with a leaner governance group should make sure a legal voice, whether in house or outside counsel, is part of the conversation given how many governance decisions carry legal implications. Stout added that governance groups benefit specifically from including a champion from the actual operational workforce, someone who can speak to how employees are really using the technology day to day rather than how leadership assumes they are using it.

Kubacki, whose Trenton based agency has built its own customized AI platform to house everything from institutional knowledge to workflows and financials, described the scale of transformation she is seeing as unprecedented, citing Microsoft data showing AI adoption driving 25 percent faster work output, 40 percent higher quality output, and 64 percent stronger innovation at even the earliest stages of implementation. At her own agency, she said productivity gains have landed around 20 to 22 percent, time her team has redirected toward higher value creative work rather than simply doing the same volume of work faster.

What It Actually Costs

Kubacki also offered a rare, concrete breakdown of what different levels of AI investment typically require, a detail many companies weighing adoption struggle to find good information on.

Douglas offered a useful historical comparison for thinking about that spending, likening the current moment to the early days of cloud computing adoption. Cloud infrastructure, he noted, did not ultimately prove valuable primarily because it was cheaper, but because it freed organizations to redirect resources toward higher value work. He suspects AI will follow a similar arc, reshaping how companies operate more than it simply cuts labor costs outright, even as the tools themselves grow more expensive over time. He framed the central business question as whether AI can finally let companies push simultaneously on speed, quality, and cost, a balance that has traditionally forced organizations to trade one against the other. Stout was careful to note that speed alone is not the goal, arguing that any efficiency gained through AI adoption is worthless if it comes at the expense of reliable, quality output.

The Cybersecurity Layer Companies Cannot Skip

Wakiyama closed the panel’s risk focused discussion with a warning about the cybersecurity implications of widespread AI adoption. He noted that malicious actors are already using AI to their own advantage, making systems harder to secure as the same tools that help legitimate businesses are increasingly weaponized for deepfakes and phishing campaigns. He described regularly fielding calls from clients after exactly this kind of incident, underscoring that companies investing in AI adoption need to simultaneously invest in their own cybersecurity posture and incident response planning rather than treating the two as separate priorities.

Panelists closed by returning to a theme that ran throughout the discussion, that many of the operational and security risks tied to AI stem from ordinary human error rather than sophisticated external attacks, making employee training every bit as important as any formal governance policy. Sailer’s closing advice was direct, urging companies to make sure managers understand exactly how and where AI is being used within their own organization, to consistently validate AI generated output, and to keep staff current on evolving internal guidance, arguing that failing to do so invites consequences that a written policy alone cannot prevent.