Top UX design agencies for complex B2B software, by situation

Insights · Article

For complex B2B software, hire the agency that works out how your records, roles and exceptions fit together before it draws a screen, and then makes every state of a busy screen easy to read. Ask every firm to show you one override, duplicate record or hand-off, both the decision behind it and how it looks to the person using it. On that standard we are our own pick for workflow products several roles share, for modernising a product expert users rely on, for dense operational screens, and for AI that suggests or acts inside shared records.

For other purchases another firm is the better call: Koru UX or Think Company for clinical or pharma software where your buyers expect a long published record in that field, Neuron for running your own design operation, Clearleft for building in-house design capability, and Fuselab Creative for a US federal programme. In each of our four situations, the table below names the firms to brief alongside us.

Disclosure: this shortlist is ours, and we are one of the eleven firms in it.

This page is for product and design leaders whose software is complex because of its roles, records, rules and history. If the hardest part of your product is onboarding or conversion in a simpler SaaS product, our comparison of eight SaaS UI/UX studios is the better place to start.

The shortlist at a glance

Read each firm's written reviews, and ask each firm for its prices, which depend on scope.

How we chose, and the standard we hold every firm to

We started from the firms that AI answers and published agency lists name for complex B2B and workflow software, then read each firm's own service and case pages. A firm is listed if it states complex product UX as a service or focus and shows at least one case on multi-role, data-dense, regulated or long-established software in enough detail to see what it decided. We described every firm, ourselves included, from its public pages. This is a starting shortlist, not a directory.

Getting on the list is the floor. In our view, complex B2B software fails between screens, not on any one of them: a scope selection that shows the wrong part of the organisation's data with full confidence, a duplicate deal nobody meant to create, a change that reaches further than the person making it could see. So ask every firm to show you these six things:

1. The product model before the screens. Does the work start from the records, roles, rules and states the software runs on? A screen designed before that model is a guess about it.
2. Exceptions and consequences beside the main path. Overrides, rejections, duplicates and holds are where customers spend their difficult days. A strong firm shows what an action will reach before someone commits to it.
3. Craft where the work is hardest. Can a person read a dense screen at first scan, and does every state, from empty and loading to blocked and overridden, look as deliberate as the main path? Interaction and visual craft decide whether expert users trust what they see.
4. Knowing what to leave alone. Expert users' habits are part of the product. Every change should earn its cost in relearning, and the firm should say why it kept what it kept.
5. Behaviour tried in code, and checked once built. Tables, drag and drop and permission states reveal their problems only with real data in a browser. Ask who tries them before the build, and who compares the shipped product with the agreed design.
6. Research with every role. Admins, approvers and whoever clears the exceptions know rules the requester forgets.

Location, review platforms and awards inform the choice and decide nothing. Outcome figures, where a firm reports them, are that firm's own claims, so we leave them to its case pages.

Workflow products several roles share

When you brief agencies on workflow software, the question that separates them is what they do where work changes hands: where an approval waits, where an exception goes, where someone retypes a record. A portfolio of finished screens does not answer it, whether your product is a first release or a suite with years of customers behind it.

Tcules (our pick)

Best for: B2B products where several roles change the same record, from a first release to an established suite, design on its own, or design carried into working code.

We begin with the records and the people who change them: who may see each record, who may change it, who approves, and which states must survive a hand-off. Only then do we draw screens, giving each role its own view of one shared record rather than its own copy of it. For Auxentios, a new quoting product for heavy industry, eight interviews mapped who contributes what to a quotation across six specialist roles; the MVP gave them one shared workflow with no fixed sequence and an interface system for a product planned to pass 200 screens (Auxentios case). On Chestnut's insurance platform we worked as an embedded design partner, in the running product and its codebase, for about thirteen months and made each consequence visible before the commitment: publishing a template names every workflow it will reach, and a payout set by hand needs a reason and leaves an audit trail (Chestnut case). A senior designer leads the work, and design engineers and developers from the same studio join as it needs them, so the thinking reaches the code without passing to another firm: on Chestnut we implemented the bonus designs ourselves and built part of the communications front end. More in B2B SaaS and workflow software.

Cieden

Cieden's enterprise focus lists internal ERPs, CRMs, HR systems, BPM and workflow platforms. Its Foqal case redesigns a support ticketing product whose Kanban board was its most-used screen: drag and drop made tickets disappear, few users could find the bulk actions, and automations could be built three different ways. Cieden describes discovery interviews, flow mapping, and prototypes tested with stakeholders and internal users in close collaboration with developers, then a reworked board and a single hub for automations.

Aufait UX

Aufait UX offers research and design for AI products, SaaS and enterprise platforms, with enterprise UX and legacy application redesign among its specialisations. Its Roca case is internal-tool work: tax and compliance notices from several business units and factory sites were tracked in individually kept Excel sheets, with no reminders and hearings missed. The redesign gives each notice an owner from registration to resolution, separates escalated notices from routine ones and takes closure to CFO approval through DocuSign, on a custom design system that replaces SharePoint's default interface.

Which to brief alongside us. Brief Cieden as well if your product is an internal ERP, CRM, HR or BPM system, the enterprise software its site lists; brief Aufait UX if your platform serves fintech, healthcare, or smart city and facility management, sectors its site names.

Modernising a product expert users already rely on

The risk in modernisation is breaking habits people have built over years: the control they use to say which part of the organisation they are in, or the way a team passes on a listing it does not want. Look for firms whose cases show what was deliberately kept, not only what was redrawn.

Tcules (our pick)

Best for: established products people already use at speed, a single product or a whole suite, where new buyers need to see a current product and existing users need the steps they know to stay where they are.

In both products here, the design started from what people already did quickly, and each change went only where it was worth the relearning. Benchmark Gensuite runs a safety and sustainability suite of more than 75 applications. We fixed the shared components first, then tried them in the scope selector, the control every application uses to know which part of the organisation a person is working in, and split it into three views: Browse, Recents and Favourites, and Nearby. We built those patterns as working prototypes in HTML, CSS and JavaScript, so Benchmark's engineers reviewed real behaviour, and a drag-and-drop question about Bootstrap 5 surfaced in design review rather than in the build (Benchmark Gensuite case). On Dealpath's deal software we redesigned one of the listing's actions: in our design, Add to pipeline looks for an existing deal before creating a new one. Marking a listing as a comparable or passing on it largely works as people already knew it, and implementation stayed with Dealpath's own team (Dealpath case). More in Product Modernisation.

Fuzzy Math

Fuzzy Math positions itself on consumer-grade design for B2B and enterprise software, and on modernising legacy systems. It has been SAP Fieldglass's design partner since 2010 on an enterprise labour management product. Its data security platform case modernises a legacy platform around admins, business executives and end users, with drag-and-drop dashboards that admins can preview as their users will see them.

UXReactor

UXReactor focuses on product insights and design for enterprise software. Its RedSeal case moved a network security company's legacy on-premises product towards the cloud over seven releases and replaced a one-size-fits-all console with eight workspaces. Its Tekion case covers ethnographic studies at car dealerships, ownership of key software modules from research to prototyping, and help building Tekion's own experience design function.

Which to brief alongside us. Brief Fuzzy Math as well if your product is labour-management software, where its record with SAP Fieldglass runs back to 2010; brief UXReactor if your product is security, networking or automotive software, the field of its partnerships with RedSeal and Tekion.

Dense tables, filters and operational consoles

Removing data does not fix a dense screen when the people using it need that data. What fixes it is a reading order built around the decision, and figures that say whether they can be trusted yet.

Tcules (our pick)

Best for: data-heavy screens where people decide or act on what they read, such as pipelines, payouts, approvals and records scoped to part of an organisation.

We design the first scan around the question the person is there to answer. Our design for Dealpath's listing card kept every field acquisitions teams rely on, regrouped into three zones so a glance shows whether an opportunity deserves a closer look. On Chestnut, a payout reads Needs Recalculation when an input has changed and Blocked when an earlier payout still awaits approval, instead of presenting every figure as settled. Tables, filters and drag and drop are where our design engineers earn their place: such behaviour is judged in the browser with real data, including the awkward cases, before your engineers commit to it (Design Engineering).

Pencil & Paper

Pencil & Paper works only in enterprise industries and publishes long analyses of enterprise data tables, filtering and error feedback that let you judge its reasoning on dense interfaces before a call. Its navigation assessment works through two examples: practice management software taken from 12 parent categories to 8, with search, a queue and favourites added, and lab software whose eight-level troubleshooting path is reworked without removing any detail. Its home page adds nine anonymised mini case studies; ask for a walkthrough of a comparable product under NDA.

Fuselab Creative

Fuselab Creative designs dashboards, data visualisation and AI interfaces for enterprise teams, government programmes and funded startups, and says senior practitioners lead every engagement, with senior designers working directly with your engineers. Its Drone Forge case brings live fleet status, airframe telemetry, weather readiness and satellite-linked command into one mission control interface. Its Fiserv Small Business Index case is a platform it designed and developed for exploring US consumer spending, with interactive charts, map-based exploration and search by sector or region. Its AI page describes consent checkpoints, reversible operations and a visible trail of what an agent did, and on Stardog Voicebox a person approves an agent's drafted action before it runs. It puts discovery at two to four weeks, followed by two-week prototype and testing cycles.

Which to brief alongside us. Brief Pencil & Paper as well, an enterprise-only practice whose published analyses of tables, filtering and error feedback set out how it approaches dense screens; brief Fuselab Creative if your product sits close to its record: UAV and drone operations (Drone Forge), public data portals such as the Fiserv index, or government programmes.

AI that suggests or acts inside shared records

Once AI starts to suggest, draft or act inside records your customers share, the agency question shifts from how the feature looks to who answers for each change: who saw it, who accepted it, and how it is undone.

Tcules (our pick)

Best for: new AI-first B2B products, and established ones adding AI suggestions, copilots or agents to workflows where a wrong change costs money, trust or a customer.

Our rule is that what an AI produces joins a shared record only once a person has seen it, and that what it changes can be traced and reversed. At Omny AI, people held the company responsible for AI text that appeared before they had chosen it; each suggestion now waits for a click, and colour marks what the AI wrote (Omny AI case). In our design for BuildTwin, where AI runs quality checks on engineering drawings, a reviewer overrides a status only by giving the corrected state and a reason, and that judgement stays visible rather than silently replacing the model's answer (BuildTwin case). We design agents that act on records by the same rule, setting what each may change and where a person signs off; our B2B SaaS and workflow software page sets out how.

We build the feature into your product with you (AI Product Engineering). For ConnectX, an AI-first building energy product, design and development ran together because the interaction model depended on what could be rendered, loaded and called: eight representative scenarios were defined, and a coded workspace showed how a plain-language question assembles the right dashboard modules and building view (ConnectX case).

Which to brief alongside us. Brief Cieden as well, whose site offers design for B2B and AI apps, including copilots and conversational interfaces. For clinical AI, Koru UX adds its long clinical record; for a US federal programme, Fuselab Creative can be bought through its GSA contract and brings a federal record.

A long public record in clinical or pharma work

When your own buyers will judge the agency on a long published record inside clinical or pharmaceutical rules, Koru UX and Think Company show more of that record in public than we do. For patient- and family-facing health products, and for health software designed and built by one team, we are the pick on our healthcare UX agency shortlist, which compares thirteen firms in more depth.

Koru UX

Koru UX positions itself around healthcare and healthcare AI, listing EHRs, clinical decision support, e-prescribing and lab information systems among the systems it designs. Its case studies include a cloud EHR platform, an urgent care EMR and a post-visit AI copilot that Koru says it built, where clinicians can reject, edit, hide or override suggestions and see the data provenance behind each one. Outside healthcare, its enterprise case covers an automation platform so complex that users needed an onboarding "university", where Koru embedded as the client's UX team alongside engineers, product managers and sales leaders. Its site presents healthcare and enterprise software as its two verticals, with an Enterprise UX section alongside UX for healthcare.

Think Company

Think Company's Merck case spans nearly a decade: research with healthcare providers, a componentised design system for Merck Connect after medical-legal review had taken weeks per page, and Merck Formulary Resources, a self-service portal for payors. There Think designed sign-on that keeps commercial and medical information apart and worked with the developers on single sign-on beneath both. It also offers embedded teams, and most of its work happens remotely.

Purchases where another firm is the better call

• Running and scaling a design operation: Neuron. Neuron designs enterprise software as a UI/UX design agency, and DesignOps is one of its three services; its services page describes engagement models from project work to full team augmentation. Its HMH case starts as design operations: workshops with 28 people across design and engineering to find the gaps and shared goals in how their teams worked. If you need help running and scaling a design operation, start here.
• Building your organisation's own design capability: Clearleft. Clearleft partners with organisations to establish design as a strategic capability, building clients' own capacity as it designs. Its Inventory Planner by Sage case describes more than three years as a strategic design partner, with a research-led project workflow now being adopted across the business.
• A US federal programme: Fuselab Creative. Federal agencies can buy Fuselab's services through its GSA contract, and it designs for Section 508 accessibility on federal projects.
• Build capacity on a settled design, or quick, low-cost screens: for that purchase, buy a development company or a production-focused team. For brand identity made alongside the product, look at the brand-led studios on the SaaS list above.

Questions that settle the choice

A shortlist narrows the field; these questions decide it. Put every one to every finalist. If your first step is an audit, ten questions to ask a UX audit agency go further.

Start with your hardest case. Give every finalist the same moment from your product, such as a record two roles both need to change, and compare the questions each asks before designing anything. With us, you can start with an optional pilot on one real workflow at a fixed price agreed up front, take on a defined project priced on its agreed scope, or keep an ongoing team led by a senior designer for a fixed monthly fee, month to month with no minimum term. How we approach the work is set out in Product Design for complex software.

Who we left out, and why

• Global innovation consultancies such as IDEO, frog and Method appear on lists that rank for these searches, including Method's own. We did not assess them here; a buyer weighing an enterprise-wide transformation programme should assess them separately.
• SaaS UI and brand studios whose published work is mostly onboarding, marketing sites or consumer flows belong on our SaaS list rather than here.

If you think a firm meets the floor and is missing, or that we have described one inaccurately, tell us and we will check it at the next review.