Résumé
Let's start with an admission that any management company executive can make without hesitation: no one on your team likes entering data into the CRM. Clicking fifteen times to update an opportunity—no one really does that.
This is the number one reason why a tool that cost a fortune remains half-empty, and why so many reports start with a data-entry catch-up session. The problem has never been a lack of willingness on the part of the teams; it’s that the interface has always been a barrier between business processes and data.
That barrier is disappearing. The American enterprise software giants have just announced, with great fanfare, the shift of their tools toward AI assistants: you won’t open the app anymore; you’ll speak to an agent who will query the pipeline, update the opportunity, and set up the meeting. Walking out of a committee meeting and dictating three sentences—everyone knows how to do that.
The phrase that sums up this shift: the interface is now AI.
We wrote as early as last November that the reign of the traditional interface was coming to an end [1]. But we need to be precise about what’s replacing it, because conversation is only one-third of the story.
The interface isn’t being replaced—it’s splitting into three distinct areas
A chatbot alone quickly reveals its limitations in an asset management firm, as we’ve documented: analysts don’t work in a question-and-answer format; they continuously compare, cross-reference, and validate information; and a chat starts from scratch, without context, forcing the user to explain everything—which creates a very real cognitive load and friction [1]. What’s emerging, therefore, isn’t “chat replacing the screen,” but a fragmentation of the interface into three complementary surfaces.
First, conversation, for one-off requests and frictionless data entry: three sentences dictated on the way out of a meeting, and the data is up to date.
Next, the HUD—a contextualized visual layer integrated into the business tool—highlights what matters at the exact moment it matters: new filings and version discrepancies in a data room, weak signals from an LP in the CRM, sensitive points in an ESG report, and a summary of what has changed before an IC [1]. The user no longer describes their context—they live within it.
Finally, the agent, working in the background: it monitors meetings, emails, and documents, and reconciles reality with the database without being asked. It is this agent that finally resolves the original sin of CRM: a database that is only as valuable as the data humans choose to enter into it.
Three interfaces, one thing in common: none of them is yesterday’s application. Your tools will indeed disappear from your screens.
The two conditions that the hype overlooks
This future is desirable, but it rests on two conditions that the wave of announcements carefully glosses over.
First condition: a true golden source. An agent who enters data into a disorganized database isn’t correcting anything—they’re hiding the bad data behind a polished conversation, until the day it shows up in an LP report that no one thought to question. Automatic reconciliation without human oversight doesn’t fix the data—it just masks it. That’s why every critical entry must be validated, and every data point linked to its precise source—down to the page and paragraph—as we already require in our HUD approach [1]. There’s no useful fundraising AI without a disciplined CRM [2]; no useful agent without a governed “golden source.” The rule hasn’t changed—it’s become vital.
Second condition: an access layer that belongs to you. Here’s the question no one asks amid the euphoria of announcements: when the assistant belongs to one American giant and the tool to another, which layer still belongs to the fund—and is accountable to it? If your agents start invoking a vendor’s objects and rules thousands of times a day, the lock-in doesn’t disappear with the interface: it becomes machine-to-machine, invisible, and more deeply entrenched. The answer isn’t to reject assistants; it’s to own the layer that acts as an intermediary: your data, your access rights, your logging, your validation rules, and your skills repository [3].
What we’re building: the Golden Source connected to the assistant of your choice
This is precisely the purpose of the MCP server we launched this summer at Bodic.
MCP, short for Model Context Protocol, is the open standard that enables AI assistants to interact with third-party systems in a structured way. In practice: your teams will be able to query and update the fund’s Golden Source from the assistant of their choice—by dictating meeting minutes, checking the status of the LP pipeline, or preparing for a committee meeting. With three guarantees that remain constant, regardless of the assistant used: access rights are those of Bodic (no access to a record means no access for the AI either), every read and write operation is logged and auditable, and sensitive writes require human validation.
A word on choosing the assistant, as it’s strategic: our MCP server is platform-agnostic. Claude, Le Chat de Mistral, ChatGPT, or whatever becomes the standard in three years. The assistant is interchangeable; the layer that belongs to you is not. This is our current, definitive take on the matter: in a world where employees will belong to an AI provider and applications to a software vendor, the only defensible position for a fund is to own the middle layer—the one where its data, rules, and strategy reside.
And this layer keeps growing. Beyond the Golden Source, we’re adding the business logic that turns it into a true operating system for the fund: a sovereign 360-degree BI platform that visualizes all this data, the fund’s marketing campaigns, secure document sharing with prospects via data rooms, tracking of LP pipelines and deal flow, generation of meeting minutes, the KYC file management engine, electronic signatures, the generation of all middle-office documents, tracking of wire transfers, document sharing with LPs, reporting of financial and ESG KPIs for portfolio companies, and even the facilitation of business communities among portfolio companies.
Every component is written in the same source code, adheres to the same permissions, and leaves the same audit trail: that is what it means to own your operating system. Intelligence integrates into your environment—never the other way around.
What Remains When the Interface Disappears
We must therefore embrace the disappearance of interfaces for what it is: excellent news—and a revelation. Excellent news, because the data entry burden disappears and your CRMs will finally be populated by agents, complete with checkpoints.
A revelation, because when the screen fades away, all that remains is what you truly own: certified data or well-hidden chaos, codified rules or verbal habits, a doctrine or reflexes. Funds that have invested in their golden source, their documented processes, and their access governance will experience the disappearance of interfaces as a liberation. The others will discover that their tool wasn’t just hiding their data. It was hiding their debt.
Your tools will disappear from your screens. What truly belongs to you, however, has never been more valuable.
References
[1] Fact Sheet #26, “Smart HUDs: The Future of AI Interfaces for Asset Management Firms,” Bodic, November 2025: https://bodic.fr/pages/podcast?slug=fiche-ia-data-hud-intelligents-l-avenir-des-interfaces-ia-pour-les-societes-de-gestion&lang=fr
[2] Fact Sheet #44, “No Useful AI Fundraising Without Disciplined CRM,” Bodic, May 2026: https://bodic.fr/pages/podcast?slug=pas-d-ia-fundraising-utile-sans-crm-discipline-ce-qui-separe-un-crm-rempli-d-un-crm-qui-drive-le-fonds&lang=fr
[3] Fact Sheet #45, “No AI Governance Without a Skills Framework,” Bodic, May 2026: https://bodic.fr/pages/podcast?slug=fiche-45-pas-de-gouvernance-ia-sans-referentiel-skills&lang=fr
If you’d like to learn more about Bodic Apps or Bodic’s consulting services: contact@bodic.eu