Something quiet but significant has already happened inside investment teams. The analyst who once opened a query builder to screen a peer group now, increasingly, opens a chat window and simply asks. The shift from menus to questions is not a future scenario to plan for. It is a working habit that has formed faster than most institutions have had time to govern.
The arrival of capable AI assistants has changed the default interface for a great deal of knowledge work, and investment analytics is no exception. What was once a specialist task, expressed through filters, parameters and exports, can now be expressed as a sentence. That is a genuine productivity gain. It is also, for any institution that takes data governance seriously, a genuine question. When an analyst asks an assistant for a Sharpe ratio or a drawdown series, where does the number come from, and can it be trusted?
This piece sets out how we think about that question, and why the answer matters more than the novelty of talking to a machine.
From menus to questions
For two decades, analytical software has competed on the depth of what it can compute and the precision of the controls it offers. That depth remains essential. What has changed is the cost of reaching it. A peer-group screen that combined several constraints once required a trained user moving through a structured interface. The same request, phrased in plain language, can now be understood, mapped to the right calculation and returned in seconds.
The benefit is not that the interface is easier. It is that the distance between a question and an answer collapses. An analyst exploring an idea can iterate in the time it takes to type, rather than the time it takes to build and rebuild a query. Comparisons, follow-ups and what-if questions become conversational. The analytical engine does not change. The way people reach it does.
The distance between a question and an answer collapses. The analytical engine does not change. The way people reach it does.
The governance question that comes with it
Convenience is not the hard part. The hard part is trust. A general-purpose AI assistant, asked for a fund statistic with no connection to a verified data source, may produce a plausible figure that is simply wrong. For an institution making allocation decisions, presenting figures to an investment committee, or answering a client RFP, a plausible-but-wrong number is worse than no number at all.
The resolution is not to keep AI out of the workflow. Analysts will use it regardless. The resolution is to ensure that when an assistant answers an analytical question, it does so by calling a trusted engine that computes on the institution's own data, rather than by improvising from training knowledge. The assistant should orchestrate. The engine should calculate. Keeping those two roles distinct is what makes the output defensible.
What institutional-grade natural language requires
- Verified source of truth. Numbers are computed on the institution's own database, not generated from a model's memory.
- A clear separation of roles. The assistant interprets and orchestrates; the analytics engine performs the calculation.
- Transparency of method. It should be visible which tool was called and which statistics were used, so an answer can be checked.
- Vendor neutrality. The capability should not lock the institution into a single AI provider whose terms may change.
Why an open protocol matters
The AI assistant market is moving quickly, and no institution should bet its analytical workflow on a single vendor remaining the leader. This is where open standards earn their place. The Model Context Protocol, or MCP, is an open standard that lets AI assistants call external tools in a consistent way. An analytics provider that exposes its capabilities as MCP tools can be reached by any assistant that supports the protocol, rather than by one favoured application.
The practical consequence is freedom of choice. A team that prefers one assistant today can switch to another tomorrow without re-engineering its analytical access. As new assistants appear, they inherit the same connection. The institution's analytical engine becomes a stable point in a fast-moving landscape, reachable through whatever interface the team chooses to use.
This is the principle behind AlternativeSoft MCP. The analytics engine our clients already rely on is exposed through the open protocol, so the same trusted calculations can be reached from the AI tools teams have already adopted. The engine and the data do not change. What changes is that they gain a natural-language front door, and that door is not owned by any single AI vendor.
What good looks like in practice
The clearest test of whether a natural-language layer is genuinely useful is whether it shortens real work without lowering standards. A few patterns are emerging among the teams adopting this approach.
Research and screening
An analyst describes the universe they want in a sentence, naming the strategy, the thresholds and the ranking. The assistant resolves the request, calls the engine and returns a shortlist. What once took an afternoon of filter-building becomes a short exchange, and the analyst spends their time on judgement rather than on operating an interface.
Due diligence and reconciliation
A document arrives claiming a particular drawdown and risk-adjusted return. Rather than taking the figures at face value, the team asks the assistant to pull the live statistics and compare. Discrepancies surface early, when they are cheapest to investigate, and the comparison rests on computed data rather than on a manual check.
Reporting and client deliverables
A portfolio manager asks for a performance and risk assessment for a quarterly review, and receives report-ready prose built on the figures the engine has just returned. The narrative is grounded in real numbers rather than reconstructed from memory, which is precisely what makes it safe to put in front of a committee or a client.
The clearest test is whether a natural-language layer shortens real work without lowering standards.
What it means for allocators
For asset owners, family offices, endowments and the teams that serve them, the strategic point is straightforward. The interface to analytics is changing, the change is already under way inside investment teams, and the institutions that benefit most will be those that channel it rather than resist it.
Channelling it means three things. First, accepting that analysts will use AI assistants and giving them a safe way to do so. Second, insisting that analytical answers are computed on verified data, with the method visible enough to be checked. Third, preserving the freedom to choose and change AI tools, so that today's decision does not become tomorrow's constraint. Handled this way, the natural-language turn is not a risk to manage so much as a capability to adopt deliberately.
The engine that institutions trust does not need to be replaced. It needs a front door wide enough for the way people now prefer to work, and a foundation solid enough that what comes through that door can be relied upon. That is the opportunity in front of the industry, and it is a practical one.