ExpertsFMAI Tools for the Investment Firm – Ben Harrison (EP.522)
Capital Allocators – Inside the Institutional Investment Industry · October 22, 2026
5 Selected Lessons:
heuristic · 2 framework · 1 reasoning_pattern · 1 interpretation · 1
Separate 'research AI' from 'firm AI' (execution on institutional memory)
Source
Capital Allocators – Inside the Institutional Investment Industry — AI Tools for the Investment Firm – Ben Harrison (EP.522) (October 22, 2026) · at 9:12 ▸ · at 5:52 ▸
When this applies
when allocating AI capability between generic knowledge-work tasks and proprietary, procedure-specific organizational execution
The takeaway
Deploy general AI for ad hoc knowledge tasks, but build or buy specialized systems that connect AI to proprietary institutional memory and codified procedures to capture durable competitive advantage.
Why it works
Frontier models are good at open-ended research and content generation, but the valuable firm-specific edge comes from AI that can draw on years of accumulated institutional data (relationships, past deals, documents) and execute the firm's unique procedures.
One level up
The defensible value of AI lies not in the model's raw capability but in what proprietary context and process it is allowed to act on.
Where it stops holding
This stops working when a firm lacks structured institutional data or codified procedures for the AI to draw on, making 'firm AI' indistinguishable from generic frontier model use.
Detail
procedure:
- Identify tasks that are generic research/analysis (decks, models, transcripts) and let frontier models handle those directly
- Identify firm-specific procedures (pitching a known LP, approaching a known management team) that depend on accumulated history
- Build agentic systems that ingest the firm's structured and unstructured institutional data
- Have agents execute the firm's specific procedure against that data to inform the next pitch or decision
rationale: The differentiator for a long-term client relationship or deal pursuit is the specific history and knowledge the firm has accumulated, not generic AI output.
In their words
Where our agentic AI lives inside of these organizations is on the execution of the firm's procedure. By that, let's say we have a long-term LP investor in our fund, or we've spent five years getting to know the management team of this company that we want to invest in.
Public-company growth pressure reframes the only unmet expectation as speed
Source
Capital Allocators – Inside the Institutional Investment Industry — AI Tools for the Investment Firm – Ben Harrison (EP.522) (October 22, 2026) · at 30:02 ▸
When this applies
when leaders of a fundamentally sound, growing business are asked to self-critique under continuous market scrutiny
The takeaway
Under continuous external performance scrutiny, reframe self-assessment around pace rather than direction, and bias planning toward the most aggressive achievable targets.
Why it works
Once a company is public, analysts and investors continuously benchmark year-over-year growth acceleration, which shifts the internal question from 'did we build the right thing' to 'did we get there fast enough,' pushing budgeting and planning to be maximally aggressive.
One level up
External measurement regimes reshape what internal leaders consider a 'miss.'
Where it stops holding
This stops working when aggressive growth targets compromise product quality or client trust, since governance-sensitive customers (as in S9) may punish rushed execution.
Detail
approach: Treat 'not fast enough' as the only real shortfall when the underlying strategy and product are sound
signals:
- Constant analyst/investor scrutiny on YoY growth
- Yearly planning sessions pushing toward most aggressive position
rationale: Once public, the goal becomes growing as fast and efficiently as possible; any gap is about pace, not whether the destination is right
In their words
If anything's not going to meet your expectation, it's that you didn't get there fast enough. We're all going to get there, but it's the question of how fast. When we do our yearly planning and budgeting, we certainly try to lean into the most aggressive position that we can.
Early AI adoption correlates with size, R&D budget, and competitiveness
Source
Capital Allocators – Inside the Institutional Investment Industry — AI Tools for the Investment Firm – Ben Harrison (EP.522) (October 22, 2026) · at 12:02 ▸
When this applies
when predicting which organizations in any industry will be early adopters of a new, resource-intensive technology
The takeaway
Expect early technology adoption to cluster among the largest, most resourced, and most competitive players in a market, with some exceptions among smaller firms that are culturally technology-forward.
Why it works
Larger, more competitive organizations have the R&D budget and internal technical capacity to invest early in unproven but valuable technology, while smaller or less competitive firms lack the resources to experiment.
One level up
Resource availability and competitive pressure, not just willingness, gate the speed of technology adoption.
Where it stops holding
This stops working when a smaller firm has an unusually strong technology-forward culture, or when a technology becomes cheap/easy enough that resource constraints no longer matter.
Detail
rule: The biggest, most competitive, fastest-growing enterprises with the most R&D resources adopt agentic AI first
threshold: top 10 firms in each vertical (PE, private credit, secondaries, law, banking) are the clear early movers
rationale: They have internal technical capacity and R&D dollars; some smaller but technologically inclined firms also move fast, so size isn't the only variable
In their words
There's a clear correlation between big enterprises that have the RD dollars, that have the internal technical capacity to do it. When we look at our client base, we look at the transition into Hentik AI capability, the largest organizations on the planet that are the most competitive, that tend to grow the fastest and have the greatest resource, those are the ones we're seeing go first on our systems today.
Vertical AI commands higher prices than vertical SaaS when it delivers real ROI, not just novelty
Source
Capital Allocators – Inside the Institutional Investment Industry — AI Tools for the Investment Firm – Ben Harrison (EP.522) (October 22, 2026) · at 32:53 ▸
When this applies
when deciding how to price a new AI-enabled version of an existing specialized software product
The takeaway
Price an AI upgrade on demonstrated client savings/ROI improvement over the prior version, not on the excitement of the technology itself.
Why it works
Because AI-driven automation produces measurably better cost savings and ROI for the client than the prior SaaS-only version, the higher price is justified by economic value delivered, not by novelty or hype.
One level up
Durable pricing power comes from quantifiable value delivered, not from technological novelty.
Where it stops holding
This stops working when the AI layer doesn't produce a measurable efficiency or savings improvement for the client, making premium pricing unjustifiable and vulnerable to competition.
Detail
rule: Charge premium prices for AI features only when they produce verifiably better savings/ROI than the prior version
rationale: We believe that that is higher value than it was before... not just because it's cool, because we think the savings and the ROI for our client is better than it was before
In their words
We believe that that is higher value than it was before. We believe that it commands superior prices from where these products have been in the past. Not just because it's cool, because we think the savings and the ROI for our client is better than it was before.
AI will compress junior-level work but likely won't replace human investment committees
Source
Capital Allocators – Inside the Institutional Investment Industry — AI Tools for the Investment Firm – Ben Harrison (EP.522) (October 22, 2026) · at 25:04 ▸ · at 0:00 ▸
When this applies
when forecasting how automation changes headcount and decision authority within any professional organization
The takeaway
Expect automation to compress time and cost at the information-processing layer of an organization well before it displaces the accountable decision-making layer.
Why it works
AI dramatically compresses the time required for junior-level information processing tasks (modeling, deck-building), producing margin/cost benefits at the junior level, but high-stakes judgment calls (investment committee decisions) still require human iteration and accountability that current AI doesn't fully replace.
One level up
Technology adoption proceeds fastest where tasks are mechanical and slowest where accountability for judgment concentrates.
Where it stops holding
This stops working if AI systems eventually gain the trust and accountability structures needed for firms to delegate final decision authority to them, which the speaker does not believe is imminent.
Detail
rationale: Clients have AI agents sitting in investment committees to see what recommendations they make, but the human committee is expected to remain the decision authority while juniors see major margin/time pickup from AI-assisted modeling and deck-building.
In their words
We've got a lot of clients who have AI agents sitting in investment committee to see what it can do and to see what recommendations it makes. Is the investment committee going to be an AI agent? I don't think so. Are they going to use it to help? For sure.