Accelerating incident investigations with AI as a co-pilot
The Challenge
Workplace safety was being undermined by fragmented tools and slow, manual processes. Teams juggled multiple systems, making investigations more complex than they needed to be. Haven Safety set out to change that by consolidating workflows into one platform, speeding up investigations and turning incident data into corrective actions.
The Approach
Haven Safety had developed a minimal viable product (MVP), and when they partnered with YLD, our role was to strengthen and refine it by bringing our product design expertise, and working closely with their team.
We began with a few focused weeks of discovery by gathering data, running workshops, and engaging directly with key stakeholders and users. This research was foundational as it clarified what Haven Safety aspired to become, surfaced the user needs that would shape our design decisions, and directly informed both the visual identity and the overall product experience.
A key design challenge was creating an AI co-pilot that balanced automation with human accountability, especially in situations that could become legally sensitive. It was essential that the AI was perceived as a supportive tool rather than a dependency that users needed to maintain full agency over their reports, with the AI keeping them meaningfully informed without ever overriding their judgement.
Designing for an AI co-pilot demanded deeper foundational thinking than a standard product engagement. User flows had to account for both human decisions and a wide range of potential AI behaviours and outcomes, which is a complexity that required careful, considered design at every stage.
Haven Safety's users also came with varying levels of technical expertise and familiarity with AI. Being sensitive to this, our design approach kept the language clear and jargon-free, and leaned on familiar behavioural and design patterns to guide users smoothly through every interaction, making the platform inclusive and accessible for everyone.
How we worked
We worked in two-week sprints with daily standups alongside Haven Safety's engineering team across the US, Europe, and India. To ensure the platform was solving real problems in real time, Haven Safety's clients were onboarded mid-project to test the platform. We incorporated their feedback directly into subsequent design sprints, refining both functionality and user experience based on what we heard.
Inside the platform and its core features
When we began working with Haven Safety, the product's goals and core functionality were already established through their existing prototype.
Drawing on our discovery research and client feedback gathered throughout , we designed and iterated on the UX and UI for five core features of Haven Safety's AI-powered incident management platform:
- Incident reporting: Users are presented with a structured, standardised form that guides them through capturing the right information from the start, reducing errors and omissions that slow investigations down the line.
- Evidence and witness collection: An AI-facilitated interview flow is embedded in the platform to conduct adaptive witness interviews, and adjust questions dynamically based on responses gathered from the user. This replaces slow, inconsistent manual processes with a consistent, thorough record.
- Incident timeline: All submitted evidence is automatically organised into a chronological timeline, with inconsistencies flagged for the investigator to review and resolve before the report can be completed.
- Root cause analysis: Users choose from structured methodologies including Five Whys, Fishbone, or Multi-Threaded analysis, while the AI surfaces patterns across the data to accelerate the investigation.
- Corrective actions: The AI capability within the platform generates actionable recommendations based on the findings. The users then decide what to accept, implement, or reject, with full visibility across the organisation on what has been acted upon.
With a product-first, user-centred approach, our work helped set Haven Safety’s platform up to evolve over time. This approach enables the platform to aggregate incident data, identify trends, clarify what happened, and help Haven Safety’s clients take informed steps to prevent future workplace incidents.
Establishing Haven Safety’s brand identity
Alongside the product design work, we identified an opportunity to help Haven Safety develop a more cohesive brand by shaping their visual and public-facing language to better reflect the strength of what they were building.
At the start of the project, their visual identity was still taking shape, their tone of voice wasn’t yet clearly defined, and their website didn’t fully reflect the strength of what they were building. Considering Haven Safety as an organisation is selling a platform that promotes workplace safety, first impressions carry real weight, and closing that gap quickly was a priority.
We started working from limited existing assets, and ran on-site stakeholder sessions to capture the Haven Safety team’s vision, identifying what the brand should look like, feel like, and stand for.
From there, we developed the full identity: logo and wordmark, colour palette, typography, tone of voice, and UI visual language, alongside a redesigned website. The tone we landed on was direct, clear, and authoritative, but approachable enough to reduce friction in high-stress situations.
The result is a coherent experience across every touchpoint, with a visual language that carries through from the public-facing website into the product itself. Haven Safety now has a brand that reflects its commitment to precision and care, signals maturity from day one, and is built to grow with them.
The Deliverables
Slashing the incident investigation timeline from months to weeks:
Haven Safety's platform built by their engineering team and shaped through our product and design partnership replaces slow, manual reporting with an AI-driven workflow, reducing friction for users, accelerating root cause analysis, and enabling smarter, faster decision-making across the organisation.
CASE FILE: The Add Evidence page has AI features integrated into it, because once the end-user submits all the relevant media files to support the incident report. The “Collect Witness Statements” section allows the end-user to provide the platform various witnesses, so that the AI can facilitate an interview with the witness. The question adapts to whatever information provides in their answers.
TIMELINE: The Incident Timeline page also has AI hugely integrated into this page so that it can utilise all the information and data inputted by the end-user from the previous pages to build up the incident report. If the AI identifies conflicting data or information, it will be flagged to be resolved, before the incident report can be completed.
ROOT CAUSE ANALYSIS: After reviewing the Incident Timeline, users proceed to the Root Cause Analysis page to uncover the underlying causes using methods such as Five Whys, Fishbone, or Multi-Threaded analysis. For straightforward incidents, the Five Whys provides a simple approach, while more complex cases can leverage any of the available methods to thoroughly articulate the scenario in their report.
CORRECTIVE ACTIONS: The Corrective Actions page is a key part of the product, where AI generates actionable suggestions to resolve incidents. With status indicators for ‘Accepted,’ ‘Implemented,’ and ‘Rejected,’ users can easily track which recommendations the organisation has acted on, helping prevent similar incidents in the future.
Closing the Engagement
Haven Safety now has an AI-powered platform that has turned an investigation workflow that once took months into one completed in weeks, enabling a 50% faster incident-to-action cycle.
Given the legally sensitive environment Haven Safety operates in, every design decision was made with that responsibility in mind, ensuring human oversight remained central to every interaction. The AI co-pilot appropriately surfaced the right insights at the right moment, reducing the stress of the investigative process while keeping the human firmly in control at every step. This approach translated directly into results, with teams seeing a 92% improvement in the speed of root cause identification.
Beyond the speed of individual investigations, the platform has contributed to a 78% reduction in repeat incidents, reflecting the broader goal of not just resolving incidents faster, but helping organisations understand and address the underlying causes to prevent them from happening again.
Before concluding the engagement, we ran hands-on workshops with Haven Safety's internal team, ensuring they walked away with the skills, systems, and documentation to own and evolve their platform independently, because a great product is only as valuable as the team empowered to run it.
View More Work
