About GaiaLens
About GaiaLens
What does GaiaLens actually do?
GaiaLens is an Applied AI consultancy that finds and captures the grounded business value of AI before you spend a pound on build. Through structured strategic consulting and design thinking, we map your data, workflows and risk exposure to identify where AI moves the P&L, then engineer secure, explainable AI to capture it. That spans data unification, regulatory reporting, document intelligence, risk detection and ESG analysis. Recent engagements have translated into 70% faster reporting cycles, 80% less manual document review, and seven-figure revenue and cost impact within 12 months.
Who is GaiaLens for?
We work with the executives accountable for the P&L, the balance sheet, the control environment and the AI agenda. That means CFOs seeking risk-adjusted returns on AI investment, CIOs who need AI embedded into the existing architecture without adding risk, COOs under pressure to cut cost-to-serve and cycle time, and Chief AI Transformation Officers tasked with moving AI from pilot to enterprise scale. If your organisation is regulated, data-heavy and cannot afford a black box, we are built for you.
How is GaiaLens different from other AI providers?
Strategy firms bring frameworks. Software vendors bring tools. GaiaLens brings both, integrated, and stays accountable for the outcome. The same team that builds your AI value case also designs, deploys and governs the model, so nothing gets lost in a handover. Every output is explainable and auditable by design, built to withstand board, audit committee and regulator scrutiny rather than to survive a lab demo.
Is GaiaLens a consultancy or a software company?
Both, in sequence. GaiaLens engages as a consultancy, scoping the problem, mapping the process and assessing the data, then builds and delivers working software the client owns. This differs from an advisory firm that hands over a strategy document, and from a software vendor selling a fixed product you must reshape your process to fit.
Did GaiaLens used to be an ESG data provider?
Yes. GaiaLens launched in 2021 as an ESG data and analytics provider for institutional investors, producing real-time sustainability scores, and won several awards for that work including Best ESG & Sustainability Analytics Platform at the ESG Investing Awards 2025. GaiaLens no longer sells ESG data as a standalone product.
That period built a specific engineering capability: real-time scoring across messy, unstructured data at scale, under regulatory scrutiny. GaiaLens builds upon this data science rich heritage to build bespoke AI solutions across a wider range of industries and use cases and still builds ESG and sustainability software where a client requires it.
Who leads GaiaLens?
GaiaLens was founded in 2021 and is led by CEO Sebastien Kirk. The team is drawn from finance, technology and academia, and the company is headquartered in London, United Kingdom.
Which industries does GaiaLens work with?
GaiaLens works with organisations where accuracy, auditability and transparency are non-negotiable. Our heritage is with financial services, (including asset managers, asset owners, banks and wealth managers), however, we work with a wide range of industries from CPG, Retail, Insurance, Utilities, Energy & Natural Resources, Public Sector, Life Sciences, Manufacturing, Telecoms etc.
How Engagement Works
How does a GaiaLens engagement work, start to finish?
A GaiaLens engagement runs through five stages: a discovery call, in-organisation process mapping, a proof of concept or pilot, a feedback and refinement round, then launch.
Discovery call. An initial conversation with everyone who needs to be involved. That means not only the project sponsor, but the people who own the data, the process and the sign-off. Getting that full group in the room early and utilising methodologies like Design Thinking, is what prevents a build being derailed later by a constraint nobody raised.
Process mapping. GaiaLens spends time inside the organisation mapping how the work actually happens and where the friction sits. This regularly surfaces a different problem from the one in the original brief, because the bottleneck is rarely where it is assumed to be.
Proof of concept or pilot. Something working, built against real data rather than a demo set. A proof of concept tests whether the approach is viable. A pilot puts a limited version in front of real users doing real work. Which one applies depends on whether the open question is technical feasibility or workflow fit.
Feedback and refinement. The build is amended in response to how people actually use it. Most of the difference between a system that works and a system that gets adopted is made at this stage.
Launch. Deployment into production, with the audit trail and governance evidence required for regulated use.
How long does it take to build a custom AI solution?
Depending on the scale of your ambition, we are fast, agile and all about delivering value quickly compared to traditional transformation projects. Typically 3 to 6 months for discovery, process mapping and a proof of concept or pilot, followed by 2 to 12 months to build and deploy, depending on complexity.
Simpler builds sit at the short end: a well-scoped document intelligence tool, or automation of a single reporting workflow. These can move faster still where scope is tightly defined and the data is already accessible. Longer timelines are driven by the state of the underlying data, the number of systems requiring integration, and the depth of governance sign-off needed before a system can go live in a regulated environment.
The front-loaded discovery is deliberate. Most enterprise AI projects that fail do so because the data work was underestimated or the output could not be explained to compliance. Both are cheaper to solve before a build than after one.
How much does a bespoke AI build cost?
GaiaLens engagements start from around £20,000 and run into seven figures for large, complex programmes. Smaller budgets typically fund a tightly scoped pilot or proof of concept against a single workflow.
Cost is driven mainly by the state of the underlying data, the number of systems requiring integration, the depth of governance and audit requirements, and whether the deliverable is a working prototype or a production system deployed into a regulated environment.
Do we need our data to be clean before we start?
No. Messy, fragmented and unstructured data is the normal starting condition, and unifying internal systems, third-party feeds and unstructured files into usable inputs is a standard part of the work. Waiting until data is "ready" tends to postpone AI projects indefinitely, because the cleaning effort only gets prioritised once a concrete use case is driving it.
Who owns the software GaiaLens builds?
The client does. GaiaLens builds bespoke software that can be white-labelled under the client's own brand, with ownership sitting with the client rather than the vendor. GaiaLens maintains it under a separate ongoing arrangement, so clients are not locked in by the licensing model. The option to bring maintenance in-house stays open.
What happens after the build, and who maintains it?
Clients choose. GaiaLens either hands over full management to the client's own team, or hosts and maintains the system on an ongoing basis. Because the client owns the software outright, handover is a genuine option rather than a theoretical exit, and clients who start with GaiaLens maintaining a system can move it in-house later without re-procuring or rebuilding.
Can GaiaLens work alongside our internal engineering team?
Yes. GaiaLens works as an extension of in-house engineering teams as well as building end to end. This suits organisations that have engineering capacity but lack specific AI or data experience, or that want a system built so their own team can maintain it from day one.
Build vs Buy, And Choosing a Partner
Should we build a custom AI solution or buy off-the-shelf software?
Buy off-the-shelf when you are comfortable with vendor lock in, paying license and maintenance forever, your process is standard, the vendor's workflow is one you can adopt without distortion, and speed matters more than fit. Build custom when the problem is shaped by your own data, your own regulatory position, or a process that is itself a competitive advantage. Build custom too when you have evaluated the market and every product solves 70% of the problem, leaving the expensive 30% untouched.
It is common knowledge that the configuration, customisation and commissioning of software is only 30% of your technology spend on a problem, so you have to ask yourself if you are happy spending more on customisation and workarounds than a purpose-built system would have cost.
What should we look for in an AI consultancy for a regulated industry?
It’s all about the industry experience, look for domain experience in your specific regulatory environment rather than generic AI credentials; the ability to demonstrate explainable, auditable outputs rather than black-box results; a clear written position on data handling and model training; delivered production systems rather than a portfolio of pilots; and clarity on IP ownership and handover before you sign.
Ask to see something running in production under real regulatory conditions and to hear about their live customers. The gap between a convincing demo and an audited production system is where most AI projects fail.
Why do so many enterprise AI projects fail to reach production?
Basically failing to understand how you as a business consume innovation successfully, including the right level of executive sponsorship and having a structured approach that measures the business value of what you are exploring and considers how it moves from pilot to go live. Starting from the technology rather than a specific, measurable workflow problem; underestimating the data engineering, which is usually the majority of the work; and building something that cannot be explained to compliance, auditors or regulators, so it never receives approval to go live.
In regulated industries the last is decisive. A system that produces a good answer it cannot evidence is unusable regardless of how accurate it is.
How is a bespoke build different from configuring a large platform vendor's AI tools?
Platform AI tools are built for the average case across a vendor's entire customer base, which means they assume a common data model and a standard workflow. A bespoke build starts from your actual data and process.
The trade-off is real. Platform tools are faster and cheaper to start and come with the vendor's ongoing roadmap. Bespoke builds fit precisely but require you to own the result. The right answer depends on whether the process in question is a commodity or a differentiator.
Can GaiaLens build ESG or sustainability software for us?
Yes. GaiaLens no longer sells ESG data as a standalone product, but builds bespoke sustainability and ESG software where clients require it, drawing on the company's original work in real-time ESG scoring, controversy detection and automated regulatory reporting. This includes sustainability reporting systems, portfolio-level analysis tooling and controversy monitoring built around a client's own data and methodology.
AI in Regulated Environments
What is explainable AI, and why does it matter in regulated industries?
Explainable AI produces outputs that can be traced back to their inputs and reasoning, rather than emerging from an opaque model. In regulated industries this is a requirement rather than a preference: compliance teams, auditors and regulators need to see how a conclusion was reached before it can be relied on. A system that cannot evidence its reasoning generally cannot be approved for production use, however accurate it is.
Can large language models be used for regulatory reporting?
Yes, but only within a controlled architecture. LLMs are effective at extracting and structuring information from unstructured documents, which is where most regulatory reporting effort is actually spent. They are unreliable when asked to produce regulated figures directly, because outputs can vary between runs and cannot be reproduced on demand.
The workable pattern is to use the model for extraction and interpretation while keeping calculation, validation and audit trail in deterministic code.
How do you stop an AI system from hallucinating in a compliance context?
By grounding every output in retrievable source documents, citing the specific source for each extracted fact, keeping numerical calculation outside the model, and building human review into the workflow at the points where the cost of error is highest. Hallucination cannot be eliminated at the model level, so systems for regulated use are designed to make unsupported outputs visible rather than to assume they will not occur.
What does an audit trail look like for an AI system?
A usable AI audit trail records what data went in, which model and version processed it, what came out, and which source documents support each output. Any individual result should be reproducible after the fact. GaiaLens builds outputs to be traceable to source, so governance and compliance teams can evidence a specific result rather than vouching for the system in general.
Is my data secure with GaiaLens?
Yes, by design rather than by exception. Our strategic consulting process begins with your data, controls and compliance requirements, so security, lineage and governance are engineered into the solution from day one. GaiaLens holds ISO 27001 certification and Cyber Essentials, and typically deploys solutions into the client's own environment rather than hosting client data centrally. Where the software runs within the client's infrastructure, client data stays inside the client's existing security perimeter and governance controls. Where GaiaLens hosts and maintains a system, it sits within GaiaLens's own certified security environment, with full audit trails on every output.
Does GaiaLens use client data to train models?
No. GaiaLens does not use client data to train its own models. Where model training is part of an engagement, it happens on the client's data for the client's own model, within their environment. The resulting model belongs to the client and is not reused across engagements.
Which AI models does GaiaLens build on?
GaiaLens is model-agnostic, selecting whichever model best fits the problem rather than committing clients to a single provider. In practice the choice is driven by the requirements of the use case: accuracy, cost, latency, and whether data residency or sensitivity rules out an external API in favour of a self-hosted open-weight model.
Self-Serve Products
What is GL Report?
GL Report is a self-serve tool that generates sustainability reports automatically from uploaded holdings, producing SFDR, TCFD, EU Taxonomy, UN SDG, SBTi, Net Zero, Sustainable Investments and Impact reports within seconds. It saves an average of 10 hours per report compared with manual preparation, and is available on a monthly subscription without a sales process.
What is GL Chat?
GL Chat is a document AI tool that lets users upload PDFs such as annual reports, corporate filings and disclosures, then query them in natural language. Instead of manually searching a several-hundred-page document, an analyst asks a direct question and receives an answer drawn from the text. GL Chat saves an estimated 240 hours of report analysis a year.
How much do GL Report and GL Chat cost, and can I trial them?
Both products can be trialled and subscribed to directly, without going through a sales process. That is a deliberate departure from the industry norm of lengthy procurement before you can see the product.
How do GL Report and GL Chat relate to GaiaLens's consultancy work?
They are productised versions of problems GaiaLens solved repeatedly in client engagements: automated regulatory report generation, and querying dense documents. Clients who need either capability embedded in their own systems, or adapted to a different framework or document type, typically engage GaiaLens to build it rather than subscribing to the standard product.
Practical
Where is GaiaLens based, and do you work with clients internationally?
GaiaLens is based in the United Kingdom and works with clients internationally, currently across the USA, EMEA and Asia. Because solutions are typically deployed into the client's own environment, data residency requirements can generally be met wherever the client operates.
Is there an API?
GaiaLens does not offer a public API into its own software. Integration runs the other way: bespoke builds connect to the client's existing systems and APIs, so the solution fits into the client's architecture rather than requiring them to extract data from a vendor platform.
How much do GL Report and GL Chat cost, and can I trial them?
Both products can be trialled and subscribed to directly, without going through a sales process. That is a deliberate departure from the industry norm of lengthy procurement before you can see the product.
How do I get in touch about a project?
Email Alex Abbott, Commercial Officer, at alex@gaialens.com, or Sebastien Kirk, CEO, at seb@gaialens.com. We will be in touch within 24 working hours.
Why The C-Suite Hires GaiaLens
AI investment with a P&L attached
You do not need another AI pilot with an undefined return. You need a business case with a number attached, and a team held to it. Our engagements start by quantifying the value at stake in cost, revenue and capital at risk, before a line of code is written. Recent results include £500k in cost savings and £1.2m in incremental revenue within 12 months, alongside a 70% reduction in time spent on regulatory reporting. The business case is not a slide; it is the KPI we are measured against.
AI that fits your architecture, not the other way round
Explainability and control are not a compliance checkbox here; they are the design brief. We integrate with what you already run, whether that is APIs, legacy databases or spreadsheets, rather than asking you to rip and replace. Every model ships with full data lineage and audit trails. That is AI you can defend to your risk committee, your regulator and your own engineering team, without adding a shadow-AI liability to your estate.
Capacity back, not just automation
We measure success in hours returned to your teams and cycle time taken out of the operating model, not in models shipped. One engagement returned 25,000 labour hours over 12 months and cut manual document review by 80%. Another substantially reduced duplicated reporting effort across teams. That is capacity you can redeploy to higher-value work, and fewer manual failure points in the control environment.
Scalable operating model, not another pilot
Your mandate is an enterprise capability with board-level accountability, not a proof of concept. We use design thinking to co-design AI with the business units that will actually run it, so adoption is built in rather than bolted on, and every solution is explainable and auditable from day one rather than retrofitted for governance later. The result is an AI portfolio you can defend to the board and scale across the business, not a graveyard of pilots.