Introducing Your New Analyst: Inferent
Over the past 12 months, I have been demonstrating techniques and workflows across frontier AI models designed to extract the full capabilities of LLMs for investment research.
The objective has always been to move well beyond simple event summaries, company background notes, or the increasingly common AI-generated stock profiles now appearing across web-based finance portals.
My focus has been on institutional-grade analysis routines that take an event, filing, transcript, data point, or information source and synthesize it into actionable investment insight in real time. Examples include full-stack earnings event analysis, sentiment and driver evolution tracking, forensic accounting checks, emerging risk synthesis, consensus forecast audits, contextual stock screening, and more.
Through this work, I have highlighted both where frontier models excel and where they fall short.
It has become clear that while models such as Claude and others are rapidly expanding their capabilities for finance use cases, there remain significant points of friction and unreliability that general-purpose AI tools cannot fully solve.
For investors, those frictions are costly. They include prompt engineering burdens, hallucinations, opaque data sourcing, inconsistent retrieval, RAG and MCP failures, constant model changes, and the unresolved issue of unregulated financial advice. The result is a persistent lack of trust in outputs, and trust is the foundation of investment research.
Financial markets are fast-moving, data-intensive, and context-dependent. They require a specialized AI platform built around verified financial data, institutional workflows, and investment reasoning.
That is the problem I am now building to solve.
The project is called Inferent.
What we are creating is your own AI Analyst.
Inferent Analyst will provide Agentic Investment Intelligence
Inferent Analyst is being designed to monitor your coverage, detect relevant market events, analyze them as they occur, and surface the implications.
Earnings releases, price-sensitive announcements, capital management, mergers and acquisitions, large price moves, management changes, shifts in analyst views, consensus revisions, macro events, and sector rotations. Inferent will be there to tell you what has happened, but more importantly, what it means.
The goal is not simply to alert investors. The goal is to deliver institutional-quality analytical insight that feeds directly into decision-making.
No more long prompts. No more missed events. No more manually assembling the same context repeatedly.
Inferent will be backed by institutional-quality data, investor relations materials, news feeds, alternative data, trusted databases, and trained analytical workflows. It will pull the right information, apply the right context, and produce outputs that investors can trust.
There are also many planned features that, to my knowledge, are not currently available across existing financial platforms. I will progressively reveal these capabilities in future posts as development continues.
Our expected launch for Inferent is the final quarter of 2026. There is still much to build, but I want to bring you along for the journey.
Each month, I will publish updates profiling the capabilities of Inferent Analyst and explaining how we are approaching the next generation of AI-driven investment research.
The feedback from finance industry insiders who have already seen the development plan has been extremely positive. I think many of you will be equally excited by what is coming.
To that end, I have created a pre-registration page. Those who pre-register for Inferent ahead of launch will receive a special bonus. You will also be able to opt in to an early-release beta testing phase.
You can pre-register here:
The Core Principles Behind Inferent
This first introduction to Inferent outlines the core principles embedded in the platform’s design.
Inferent Analyst is intended to be fundamentally different from traditional data terminals. Existing terminals provide investors with screens of numbers, charts, estimates, and news. They are powerful data access tools, but in reality they just provide data overload as they do not communicate any real insights or conclusions.
Every investor knows the feeling of spending hours reviewing financials, estimates, transcripts, charts, and industry data, only to arrive at the same question:
“So what does that all really mean for the share price?”
Inferent is being built to help close that gap.
Core Principle 1: Lean into AI’s real strengths
Too many AI deployments in finance are simply natural language interfaces placed on top of financial databases.
That may sound innovative, but in practice it often creates more friction and as a business model will not last. Typing a chat message to retrieve a data point is frequently slower than clicking through a terminal. If AI is not being used to understand context, reason through information, and synthesize insight, then it is little more than another search layer, often with more friction than traditional tools
It may be novel at first, but user experience shows that investors quickly revert to the faster tools they already know. How many of you have tried AlphaSense or even a Factset MCP connected to Claude and are still using it regularly for bespoke research? Many conversations I have had with professional investors reveal the same frustrations with these sorts of deployments and result in rapid user atrophy.
AI’s real strength lies in its ability to process contextual information alongside quantitative information. For investors, this means that documents such as earnings releases, presentations, transcripts, management commentary, broker research, news, and macro releases can finally be analyzed together with financial metrics to create a more complete picture.
This is where AI can excel at replicating elements of the traditional equity research functions.
Inferent is being designed to make investors faster by removing friction, not adding it. It will also treat context as a core analytical input alongside conventional financial data, leaning into AI’s real strengths.
Core Principle 2: Analysis should accompany all data
Inferent is being built as an analysis-first platform.
The familiar data will still be there: stock price movements, consensus estimates, estimate revisions, analyst recommendations, adjusted financials, news, transcripts, and more.
But the data will not sit alone.
Wherever financial or contextual data is presented, Inferent will also provide a current analysis of what that data implies. The platform is designed to connect the numbers to the narrative, and the narrative to the investment decision.
No more screens of numbers without interpretation.
Core Principle 3: Real artificial intelligence should be autonomous.
Markets move quickly. Typing chat messages is slow.
When earnings hit the tape or an unexpected event occurs, investors need to understand the implications immediately. They should not have to write a long structured prompt, manually collect consensus estimates, retrieve prior-period context, and hope the model uses reliable sources.
Long prompting will not be the end-state for AI in investment research.
Inferent Analyst is being designed to be agentic wherever possible. Define your coverage universe, and the system will monitor events, identify what matters, analyze the information, and present the insights automatically.
When bespoke analysis is required, users will be able to access and deploy institutional-grade analytical workflows (and create their own) with a few clicks, rather than writing a long prompt from scratch.
Core Principle 4: A specialist platform can solve what generalist models cannot
In my 2026 “Best Practice in AI for Investment Research” survey, professional and individual investors both identified ongoing frustrations with AI.
Those frustrations are not minor workflow issues. They sit at the heart of whether investors can trust AI-generated research.
Below are three of the largest issues Inferent is being built to solve.
Investor Frustration 1: Lack of trust in outputs
Hallucinations remain a major problem.
Even in 2026, investors still encounter model outputs that misstate facts, misrepresent dates, fabricate sources, overstate the number of websites searched, rely on low-quality sources, or produce citations that do not support the claim being made.
For investment research, this is unacceptable. My latest proof of this problem and discussion of why it is occurring in one of the world’s top frontier models is discussed here.
A confidently presented but factually incorrect output can break an entire investment argument. Investors are then forced to spend time checking every statement, which defeats the purpose of using AI to accelerate research.
The problem is amplified by the nature of financial data. Company data is fragmented across filings, IR materials, news articles, estimate providers, databases, market data feeds, and third-party platforms. Metrics are often adjusted, standardized, restated, or defined differently across sources.
This noisy and dispersed data environment distracts the model’s reasoning capability and creates unreliable outputs.
Inferent’s solution
Hallucinations build when models can’t identify the exact data they need to answer a query or are prevented from doing so. Inferent will not rely on open-ended web search as the foundation for filling knowledge gaps to produce investment analysis.
Instead, it will use native, first-party data and Inferent’s own data structures and internal models to anchor its outputs. The aim is to create an environment where analysis is grounded in verified financial, market, and contextual data from the outset within a single system.
The platform is being built to make source reliability a feature of the system, not an after-the-fact user responsibility.
Investor Frustration 2: Prompting friction
Prompting remains slow, frustrating, and inconsistent.
Better prompting produces better analysis, but most investors do not have the time to repeatedly design, test, and refine highly specified prompts, particularly in live market conditions.
Models are improving through pre-loaded “skills” and finance-oriented workflows. A general prompt such as “analyze Microsoft’s earnings release” can now produce a reasonable output.
But reasonable is not enough.
Sophisticated investment research often requires more than a generic output. Investors may want sentiment scoring, driver evolution, consensus bridge analysis, segment-level interpretation, accounting quality checks, peer comparisons, management language shifts, capital allocation analysis, or industry-specific diagnostics.
That still requires a highly structured prompt, populated with reliable data and carefully specified instructions. In a fast-moving market, this is not a scalable workflow.
Inferent’s solution
Inferent is being designed so that prompts are only required when a task cannot be pre-specified or triggered automatically.
The platform will come pre-packaged with sophisticated analytical workflows based on the research routines already demonstrated through Inferential Investor.
These workflows will operate in two ways. First, they will run agentically when specific market events trigger them - define your coverage and analysis will be presented to you as developments occur.
Second, when users want to conduct bespoke research, they will be able to deploy specialized workflows, including industry-specific analysis and their own custom designed tasks, with a click rather than a prompt.
Users will also be able to build and store their own custom workflows inside the Inferent environment. Those workflows can then be reused, scheduled, or deployed automatically.
In other words, investors can retain the flexibility and creativity of frontier models while gaining a more reliable, repeatable, and investment-specific workflow.
Investor Frustration 3: Up to date financial and contextual data
Current data is central to investment research. Yet reliable data access remains one of the largest frustrations for investors using AI.
At their core, AI models are historically biased. They are trained on past information that has already been written, published, and absorbed into their parameters. They have a tendency to over-weight this information due to its parameterization in their architecture and this affects analysis and conclusions. When they retrieve current information, they rely on external tools, browsing, APIs, plugins, or MCP connections. In doing this via external RAG to other systems or the web, research has shown it distracts the model, particularly where the data is noisy or very fine distinctions separate similar variables (like normalized vs reported). When a model gets confused, it falls back to its training data which may be years old and its reasoning skills are interrupted.
These, therefore, are often imperfect integrations.
Anyone who has experienced a failed data call knows the issue. The model may not retrieve the correct period, metric, adjustment basis, data format, or source. It may confuse quarterly and annual data. It may fail to distinguish reported, adjusted, normalized, or standardized figures. It may not understand the naming conventions inside the database being queried.
The user is then forced to specify and then re-check every parameter manually.
What period? What line item? What source? What metric definition? What adjustment basis? What frequency? What output format? What endpoint?
This is not an efficient solution.
There is also the cost issue. MCPs and data connectors are useful only if the investor already has access to expensive data infrastructure. Many investors are not going to pay for a full FactSet subscription on top of their AI subscription and token costs.
Inferent’s solution
Inferent will maintain its own data sources, APIs, and databases.
The platform will be trained around those data structures so that current, verified financial and contextual data sits at the core of every analysis. This includes earnings releases, transcripts, market data, news, consensus information, and other relevant sources.
Even custom workflows will be able to link directly to Inferent’s data fields. Users will be able to specify time periods, intervals, metric definitions, and output formats inside the Inferent environment with a few simple clicks, rather than writing detailed data instructions into a prompt each time.
The objective is simple: reliable, repeatable, current-data-driven analysis.
Inferent Analyst will be different
Inferent Analyst is being built to be very different from what exists today.
It is not just a data terminal.
It is not just a chatbot.
It is not just a document search layer.
It is an attempt to build agentic investment intelligence around verified data, institutional workflows, contextual reasoning, and automated insight generation. It is your Analyst.
I am excited to bring you along as the platform develops. Each month, I will introduce a new concept in agentic intelligence for investors and show how Inferent is being built to change the investment research workflow. I can promise that these updates will introduce you to some very novel ideas that you have not yet seen in the marketplace.
Stay tuned, pre-register, and share this post with others who may be interested. And as always let me know your thoughts…
Andy West
The Inferential Investor
