What the Second Half-Hour Is Worth
Most conversations about artificial intelligence in professional services begin with the wrong question. The question they start with revolves around how much time the person, the department, or the company saves, but what really matters is what you or the department or the company does with all the time savings. These savings can end up somewhere - in a bank of flex time or by digging deeper on an issue. How these questions are considered and answered ultimately determines whether AI adoption produces better work or merely just allows more mediocre work to proliferate. If you’re a practitioner in a fiduciary role - that difference is consequential.
Charlie Douglas, JD, CFP, CAP, AEP, an independent trustee and family office advisor, has made a career our of working alongside families from both the institutional perspective and inside a single family office. In a conversation with The Family Office Project, he laid out his role transitioned from head of a family office to one of their trustees along with the ways he’s incorporating AI use to dig deeper into his role for the family.
The Productivity Claim That Understates the Case
Charlie Douglas is an attorney, certified financial planner, and trustee in a single family office. In a recent Family Office Insights conversation, he offered a direct answer to the efficiency question: AI reduces his working time on most projects by at least fifty percent. That figure tracks with what practitioners across disciplines are beginning to report, and it tends to be the number people focus on, but Douglas pointed to something more structurally interesting. When a task that would have consumed an hour is completed in thirty minutes, the remaining thirty minutes do not disappear, they become available. The question is how a practitioner chooses to use those next thirty minutes.
Douglas's answer: he uses that time to go further into the same project. For instance, he tests additional scenarios or considers angles he would not have reached while working against a manual deadline. The deliverable that would have been adequate -- completed, acceptable, submitted -- becomes a more developed and considered answer. This doesn’t happen by virtue of using AI to improve his judgement, but because he used newly freed time to apply deeper thought processes to his workflows.
The value proposition changes from efficiency to depth.
Why the Gain Is Not Evenly Distributed
The depth argument depends on a precondition that is easy to overlook: the practitioner must know what to ask next.
While AI can compress the time required to produce a first draft, a summary, a SWOT analysis, or a negotiating framework, it cannot determine which subsequent questions are worth asking. That determination belongs to the practitioner, and it requires the kind of domain knowledge that accumulates over years of direct work -- not general familiarity with a field, but specific fluency in its patterns, its failure modes, and its edge cases. This is in direct conflict with the techno-futurist idea (or hope?) that AI will eradicate entire incoming entry-level cohorts of new hires. Without the accrued daily, weekly, yearly education on the job or in the field, no practitioner will have the opportunity to become a domain expert.
Douglas also makes a point about the quality of AI output and it’s identifiable characteristics. When someone submits AI-generated work without applying substantive judgment to it or even editing, the gap is apparent to those with enough experience to recognize it. The floor for useful AI adoption is not the adoption itself but the informed adoption -- using AI as a starting point that experienced judgment then improves.
Practitioners who use AI to avoid thinking produce faster garbage, whereas practitioners who use AI to think further produce better work. The difference between those two outcomes is entirely a function of what level of domain expertise the practitioner brings to the process before the first prompt is typed.
What This Means for the Trustee Role
The fiduciary context makes these distinctions more precise. Douglas described how he uses AI across the three core functions of trusteeship: investment oversight, trust administration, and distribution. In practice, that means running portfolio analyses, negotiating real estate transactions, drafting beneficiary communications, and building materials that explain trust structures -- including the tax implications of grantor vs. non-grantor arrangements -- to family members encountering them for the first time.
None of these use cases involve delegating professional conclusions to AI. Douglas was explicit: he does not use AI to render legal or tax advice. That responsibility stays with counsel and accountants. AI handles the preparation and the drafting; human judgment handles the conclusions.
He also described a data discipline that most offices have not formalized: redacting account numbers and identifying information before uploading documents, using snippets rather than full statements, and cross-checking outputs across multiple platforms before relying on them. These are not informal habits. They are governance decisions operating without a governance framework.
The absence of that framework is itself a structural risk. AI adoption in family offices is already underway -- formally in some offices, informally in others. The offices that have defined their standards around data handling, output verification, and the boundary between AI-assisted thinking and AI-substituted judgment are better positioned than those that have not.
The Competency That Is Redefining Entry-Level Expectations
Douglas invoked a colleague's framing that is worth taking seriously as a governance principle rather than a motivational observation: practitioners who learn to work with AI will have jobs; those who refuse will be replaced by those who do.
This is not his prediction about mass displacement. It is a narrower and more immediate claim: within professional roles that already require domain expertise, AI competency is becoming a de minimus or threshold skill. The practitioner who can prompt well, evaluate output, identify hallucinations, and use recovered time to go deeper is producing better work than the practitioner who cannot. Over time, that gap compounds, just like domain expertise.
For family offices, the implication is structural. AI governance is not only a question of data security protocols and platform selection. It is also a question of whether the office is developing the practitioner capability required to use the technology well.
What the Evidence Suggests
The practitioners who are getting the most out of AI are not the ones treating it strictly as a search engine or a drafting shortcut, but the ones treating it as a collaborator whose output requires evaluation. The catch is, much like an entry-level employee, it can produce work that is fast, broad, and can be wrong in ways only experienced judgement can catch. That is the specific model of education or training you see everywhere in industries, experience passing down to younger generations - and now for many - the younger generation also includes AI models embedded within the team.
For family offices navigating governance, technology selection, or succession planning, explore the Family Office Project on YouTube or connect with Jon Carroll + Family.
