Bringing the document library to where brokers already work.

Marsh is one of the world's largest insurance brokers and risk advisers. A broker's day runs on documents: quotes and proposals, binders, endorsements, invoices, premium schedules, and the email trails that go with every placement. For a single client, that can be hundreds of files across many policy years.

MDocs was Marsh's document store. It worked, but it sat apart from the systems brokers actually place and service business in: MART, Placement Manager, and the Colleague Desktop. Filing meant switching context; finding meant remembering where something had been put.

The brief was to design a Doc & Email Library that could be reached from inside those systems, so a broker never has to break their flow to store, retrieve, or share a file.

MDocs did its job as a store, but the path to it was the problem. Every interaction pulled the broker away from the task in hand and asked them to do the librarian's work by hand.

What made it slow

A separate destination. Opening MDocs and uploading documents was a step apart from the work that produced them.
Log in, then hunt. Getting to the correct client and folder was a repetitive, time-taking process every single time.
Hard to keep current. Tracking and updating files as they changed was difficult, so versions drifted.
A poor filing system. For a broker juggling many clients and policy years, the folder structure did not match how the work is actually organised.
No bulk. Files could not be downloaded in bulk, so pulling a set of documents meant repeating the same manual steps.

The tell came straight from the people who used it: "I want to decrease my time and effort in arranging documents." When the goal a user states is simply to spend less time fighting the tool, the tool is the work, not the help.

The crux of the solution came down to three promises to the broker, plus the one structural bet that makes all three possible: put the library inside the work.

I led the UX design for the Doc & Email Library: framing the problem with the client, structuring the information architecture, mapping the task flows, and designing the interactions and visuals from first sketch to the validated hi-fi screens.

I owned the discovery and a UX expert audit of the existing MDocs experience, the research synthesis, personas, user and task flows, the information architecture, low-to-high fidelity wireframes, and the visual design through to usability testing. I worked alongside Marsh business and product stakeholders for direction and sign-off, broker subject-matter experts for context and validation, and engineering for feasibility and build.

Discovery ran on three tracks: understand the people, analyse the problem, and develop the concept. I collected data on how documents are used and did a behavioural assessment of daily tasks, ran a UX expert audit of MDocs alongside competitive analysis and benchmarking, and worked up early directions through brainstorming and group discussions with stakeholders.

What I heard, again and again:

"I spend too long just getting to the right client and file." · "I want to upload several documents at once." · "After I upload, I am not sure it actually went through." · "Keeping every file current across policy years is hard."

A note on honesty: the personas and journey that follow are design personas, synthesised from the user research, the expert audit, and stakeholder conversations. They are decision-making tools, framed as such, not a claim of large-sample quantitative study.

The heart of the product is a single, dense document list. Four tabs, Account, Placement, Policy, and Billing, mirror how broking work is structured, so a broker lands exactly where the document belongs. Actions that used to be separate trips now sit on the same screen.

The single biggest time sink was gone: no more logging in and walking a folder tree. A collapsible filter rail narrows by subtype, format, product line, coverage, and status, while a search field under every column handles the rest. Look things up the way brokers think, by name, date, carrier, or status.

Efficient search, delivered. Dynamic searching and filtering replace the folder hunt. The document a broker needs is a keystroke away, not a navigation task.

Adding a document used to mean a separate upload chore. Now it is a short, guided flow that starts from the list itself: drag a file in, let the system suggest its name, confirm a few details, and it lands in the right place. Documents and emails can come straight from Outlook, a shared drive, or the desktop.

One of the loudest complaints was uncertainty after an upload. A notification centre now groups every upload into three clear states, so a broker can see what went through, what is still processing, and what needs another try, with a one-tap re-submit on anything that failed.

The visual system leans on Marsh's blues and a restrained, data-dense layout. The point of a document tool is the documents, so the interface stays quiet and lets the list do the work.

Density with air. A high-information table, but generous rows and clear column heads keep it scannable.
Search on every column. A search field sits under each header, so filtering is where the eye already is.
Status is icon plus colour. Success, in-progress, and failed each carry a shape, not colour alone.

The engagement ran on a three-step user-centred method: understand the work deeply, design a viable solution against it, then prove the solution with the people who do the work. Structure was settled in low fidelity first, then taken to build-ready hi-fi, testing and iterating along the way.

The final E, Examine, was not a formality. The design went through usability testing and multiple rounds of iteration against clear metrics, pressure-testing every scenario, bottleneck, and breakage before it was signed off.

The engagement was framed around three outcomes for Marsh and its brokers. Each maps directly to a decision in the design. With no live post-rollout figures to hand, the honest read is the intended value, plus the signals I would instrument.

If I did it again

Instrument the live product against the signals above from day one. Push automatic naming further, learning conventions from how each team actually files. Test the embedded entry points with brokers in their real workflow, not just the library in isolation.