The AI coding revolution.
If you work in investment management and have spent any time with Cursor, Claude Code, or OpenAI Codex in the past six months, you’ve probably had the thought: We could just build this ourselves. AI coding tools have dramatically lowered the barrier to getting a prototype running. That’s genuinely exciting.
But a prototype is not a product. And a product is not a platform your team will rely on every day to make investment decisions worth billions of dollars.
We say this not as detached observers, but as builders who live the build-vs-buy tension every single day.
We Buy and Build
At Calibre, we are voracious buyers of software components. Our research management system integrates services and tools from numerous vendors, large and small, covering everything from AI model providers (Azure OpenAI, Anthropic, Google, AWS) to data feeds, authentication, cloud infrastructure, UI components and more. We don’t build our own large language models. We don’t build our own SSO identity layer. We don’t build our own rich text formatters or consensus data feeds.
Why? Because like every company, we have scarce resources. We want to allocate product roadmap and engineering time where we can create the most value for our users: building the best research management system for institutional investment teams. That means focusing on the workflows, governance, and integration points that are unique to how analysts and portfolio managers actually work, and buying commodity components from specialists who do them better than we ever could.
We know exactly what building means.
We believe this is the same calculus every investment firm should apply when considering whether to build an internal RMS.
What Looks Simple Isn’t
The initial build is intoxicating. You stand up a database, wire in an AI API, generate some structured notes in a slick web front end, and within a week you have a demo that impresses the CIO. We get it. We’ve been there.
But then the real requirements emerge:
Institutional-grade MS Office integration: Microsoft add-ins, Excel modelling with check-out/in, version history, decimal point accurate calculation and normalisation scripts, distribution permissions, published time-series, and consensus vs. internal side-by-side diffs. This is not a weekend project.
Permissioned, auditable AI: role-based access, model selection governance, BYO API keys, system prompts, grounded citations from source documents, protection against the LLM lethal trifecta, and zero data training commitments from all underlying providers.
Regulatory compliance and record-keeping: comprehensive audit logs across every action in the system to comply with s1043A Corporations Act 2001 (Australia), FCA’s SYSC6.1 and 4.1.1R (UK) and SEC Rules 17a-3 and 17a-4 (US).
ESG and stewardship workflows: custom scorecards, engagement logs, proxy voting record integrations, regulatory reporting for SFDR, TCFD, and stewardship codes, all linked back to the investment case.
Decision logs and process workflows: investment thesis tracking, bias checks, post-mortems, analyst checklists, and portfolio-level pass/fail process tests that update in real time and fit naturally into analyst and PM workflows.
Multi asset class support: handling the securities mapping, data fragmentation, differing analytical frameworks, and varied ESG considerations across equities, fixed income, alternatives, and real assets.
Security certifications and due diligence: SOC 2 and ISO 27001 certifications require rigorous, ongoing third-party audits of security controls, data handling, and operational processes. When you buy from a certified vendor, you effectively outsource the non-core-competency due diligence burden. Building in-house means owning this certification burden yourself, or asking your stakeholders to accept unaudited assurances.
Long-term maintenance: version upgrades, security patches, regulatory changes, database schema evolution, cloud infrastructure management, commercial data licensing & integration, changes to accounting standards, new LLM model capabilities, onboarding new team members, surviving the departure of the engineer who built it.
These are some, but not all, of the realities to consider. Each of these is a layer of complexity that compounds. The build that took two weeks to prototype takes six to twelve months to make production-grade (not including time negotiating with internal corporate IT informing you what you can and can’t do). Then it needs a team to support the users, fix bugs and maintain it indefinitely. Internal builds often create key-person risk and typically lack the data quality, governance, analytics, user support and automation that a comprehensive platform requires.
Tactical Fit, Strategic Flexibility and Evolution of Requirements
There’s another dimension to the build-vs-buy decision that becomes apparent only with time: the difference between tactical and strategic systems.
In-house builds are typically optimised for today’s requirements. They’re tactically fit for purpose, solving the exact problems the team faces right now, with the exact workflow they currently use. This feels efficient. But investment environments change. Strategies evolve. Teams restructure. Regulatory requirements shift. Asset classes are added or removed.
What was perfectly tailored becomes brittle.
CalibreRMS was built over more than two decades with a different philosophy: configurability and extensibility as core architectural principles. Many of the use cases Calibre supports today, such as AI-assisted research, ESG integration, stewardship workflows, multi-asset class coverage, were never even contemplated when the platform’s foundations were laid. They were possible because the system was designed for flexibility rather than for a fixed specification or a single problem.
This is why Calibre can incorporate AI capabilities in ways that feel native rather than bolted-on. Our approach to implementing each team’s process has always been driven by a philosophy of flexibility: configurable workflows, extensible data models, and an architecture that assumes requirements will change.
We Encourage You to Build. Seriously
Despite all of this, we genuinely encourage investment teams to consider building. Our observation, after working with institutional investors globally, is that software is dramatically under-deployed in investment decision making. If the AI coding revolution gets more teams excited about using technology to sharpen their process, that is unambiguously a good thing.
But here is the question worth sitting with: Are you rebuilding a commodity tool (and maybe saving a few dollars) or are you using your newfound enthusiasm for software to build something genuinely different that provides you a unique investment edge?
A research management system is a coordination layer. It needs to work for every analyst, every PM, across every company, instrument and asset class your team covers, every single day. It is like infrastructure for investment decision making. Building your own version of infrastructure that already exists is a bit like an investment firm building its own Bloomberg terminal – technically possible, but a peculiar use of resources.
We believe where the real opportunity lies is in building on top of that infrastructure. Custom AI models, proprietary data integrations, unique analytical frameworks, bespoke screening tools. These are features no platform provider will build for a user base of one.
But these are the features that by definition can give your team a genuine edge. And they’re far easier to build when you’re not also maintaining the plumbing underneath.
Before You Start, Talk to Us
For those investment teams determined to build, we are more than happy to help. Reach out to us and we will provide a checklist of everything you should consider before you embark on this journey – from data architecture and security to workflow design and regulatory requirements. No sales pitch required. We have learned these lessons the hard way over more than two decades building CalibreRMS, and we would rather you go in with eyes open.
CalibreRMS offers API and MCP server access to the full CalibreRMS system, which means if you do build your own technology, you can pull data from CalibreRMS and/or publish back into the system. Our philosophy is that the research that lives in CalibreRMS is yours, and you should be able to access it and do with it whatever you want, without us putting barriers in your way. Platforms which try to capture users rather than allow them to extend are those most likely to be built around or displaced.
As daily users of AI tools ourselves, we wish they were a silver bullet for product design, software development and infrastructure maintenance. Maybe that day will come soon. The hardest part of building an investment research management system was never writing the code, it was knowing what to build, and having the discipline to maintain and grow it for the long term.
Build where it gives you edge. Buy where it gives you leverage. And whatever you do, use MORE software in your investment process.
——————————
Click Here for Interactive Demos:
Share on Facebook
Share on X
Share on LinkedIn
