Project highlights
Task
Build an AI-native internal platform to unify fragmented operations across sales, delivery, finance, and team workflows.
Team
Team: Yuriy, CEO & co-founder (engineering lead, part-time), with design support added later.
Duration
12 months, part-time.
Scope
Sales, delivery, finance, legal, staffing, reporting, AI-assisted workflows.
About the client
Cieden is an AI-native product design company that helps founders and product teams design and ship complex digital products. With 50+ people managing international client engagements, diverse internal teams, and intricate project lifecycles, we needed an internal business operations platform that matched the standard we hold for our clients' work.
What the client needed
As we grew, operational complexity began to outpace the tools holding it together. The core problem wasn't any single tool — it was that we had no internal crm connecting all the parts together. Fragmented tools, disconnected workflows, and weak handoffs between departments meant every team was working from a different version of the truth.
What we needed:
-
One source of truth connecting sales, delivery, staffing, finance, and legal
-
Automatic links between sales opportunities and resource planning — no manual handoff from contract to execution
-
Real-time visibility into invoices, contracts, and payment statuses for finance and legal teams
-
A pipeline built for forecast accuracy and strategic decisions, not just a list of leads
-
HR and utilization data alongside delivery data — the full picture without switching tools
-
AI-native workflows built into daily operations, not bolted onto legacy process logic
Existing CRMs and off-the-shelf tools each solved one part of this. The limitations of legacy systems became clear as we scaled — none of them covered the whole thing.
Challenges
- Syncing sales opportunities with invoices, projects, and staffing in real time — across what had been entirely separate systems
- Building role-based access for 14 distinct roles, from Legal to Design, without creating an ongoing permissions burden as we scaled
- Preventing data duplication when multiple departments entered data concurrently into the same shared model
- Making pipeline stages, staffing gaps, and team utilization visible at a glance — without manual report assembly
- Scaling operations without introducing new chaos: the system had to grow with the team, not create architectural debt
- Building for ourselves: our colleagues were the users, their daily work depended on it being right, and requirements kept shifting as real usage revealed what actually mattered
Results
- 58+ active team members migrated from fragmented apps into one shared platform
- 30% faster project onboarding by linking signed contracts directly to team allocation
- Manual sync between CRM and invoicing eliminated — automated data flow replaced disconnected spreadsheets
- Staffing and hiring needs forecastable 2–3 months ahead of project kick-offs
- Improved cross-team visibility: sales, delivery, and finance now operate from the same data
- Approval workflow software replacing email-based approvals — finance and legal reviews now run inside the system
- Legal workflow automation covering contract tracking, compliance exports, and approval histories
- 14 deep external integrations including HubSpot, Zoho Books, Concord, Upwork, and n8n
- 70+ schema tables, 200+ indexes, 8+ real-time webhooks, 15 automated jobs
We started with a product discovery phase — a deep audit of where our process actually broke down, not where we assumed it did. The finding was consistent: friction lived in the handoffs between tools, not inside any single one. Data that should have flowed automatically was being manually copied.
One principle governed everything from that point: the platform had to be relational. A deal had to connect to the project it became, the team allocated to it, the invoice it generated, and the contract governing it — automatically, not through someone in the middle updating four systems. Unlike most CRMs where data goes stale the moment it loads, we built on a real-time backend where every change propagates across all modules in under 100ms.
The crm system design ran across three parallel tracks: a relational data model linking opportunities to invoices and staffing; operations dashboard keeping complex data readable at scale; and automated workflows removing manual bottlenecks from finance and legal approvals.
We used the platform in real work from early stages and gave continuous feedback. When something didn't work, it changed. The platform is still evolving today.
It's the same thinking we bring to our CRM design services, when working with clients on similar operational challenges.
Sales and forecasting
We tracked deals in HubSpot and built forecasts in spreadsheets — two separate exercises that never talked to each other. HubSpot told us where deals were; it had no view of where revenue was heading, which deals were going cold, or how the pipeline connected to next quarter's staffing needs. Analysis lived elsewhere, which meant it was always behind.
We rebuilt the sales module around a filter-aware KPI header that recalculates weighted forecasts, untapped pipeline value, and lead response times the moment any filter is applied — turning the pipeline board into a live what-if analysis tool with urgency signals when thresholds are exceeded.
Nine analytics views go deeper: Win/Loss analysis, ICP effectiveness, service-line revenue contribution, and channel-level LTV. Sankey flow visualizations track stage-to-stage conversion interactively. Average Cycle Time, Deal Velocity, and Time in Stage are visible in real time.
Sales teams manage deals while leadership inspects where momentum is building or breaking down — all from one sales CRM dashboard, without waiting for reports. This kind of clarity is what our dashboard design services are built around.
Resource planning and delivery
We built a resource planning dashboard that connects pipeline opportunities directly to team capacity. Each opportunity carries budget, timeline, and staffing ratios. Utilization heatmaps surface delivery bottlenecks before they impact projects. Time-off data integrates into capacity calculations — leave balances, upcoming absences, PTO trends — so nothing appears as a surprise.
Project resource planning is now proactive — staffing gaps are visible 2–3 months ahead. Hiring decisions are grounded in live resource forecasting data. We stopped being surprised by delivery pressure.
Finance, legal, and approvals
Before the platform, invoices lived in one tool, contracts in another, and approvals happened over email. Finance knew what had been invoiced. Legal knew what had been signed. Neither was connected to delivery or to each other. Tools like QuickBooks handle accounting and DocuSign handles signatures — but neither understands the project behind the invoice or the pipeline opportunity that preceded the contract. Reconciling across them was a manual exercise that repeated every month.
We unified invoices from Upwork, Zoho Books, Avaza, and manual creation into one pipeline with automated status tracking. Finance workflow automation handles overdue detection and aging analysis. Commission tracking covers 8 role types with 12-month attribution windows. On the legal side, AI automatically extracts 20+ structured data points from signed contract PDFs — liability caps, termination periods, confidentiality terms, billing rates — and scores every agreement for risk from 0 to 100. Alerts fire at 90, 30, and 7 days before expiration. Approval histories and compliance exports are available on demand.
Finance and legal reviews now happen inside the platform — connected to the projects they relate to, not isolated in separate tools.
AI assistant inside operations
Even with everything in one platform, getting the right information before a client call still meant opening four modules: deal history, team allocation, contract status, recent invoices. Most internal tools have no real AI layer, or add a chatbot that doesn't understand the actual data. An AI assistant for business that can't query real operational data isn't useful for daily work — it's a search bar with a different interface.
We built a Claude-powered agent with 24 CRM-aware tools, accessible from anywhere via Cmd+K. It understands full operational context across all 18 modules — contacts, deals, projects, invoices, agreements, team capacity, deadlines — and routes queries in real time: direct lookups go straight to data, complex questions go through full AI reasoning, all at sub-100ms latency. The assistant also generates structured proposal drafts from live CRM context: approach, pricing, timeline, and team in one action.
This is what makes the platform AI-native rather than AI-adjacent — intelligence woven into how the team works, not added on top. It's also how we think about AI as an AI design agency: not a feature bolted onto a finished product, but a layer built into the logic from the start.
Building this for ourselves was the real test — no client buffer, no room to hide. The platform had to earn its place in daily work. It did.
If operational complexity across disconnected tools is a problem you recognize — we've been there. As an AI design agency that builds products like this for clients, we bring the same thinking we applied here. See more case studies from our work.
we reply under 24 hours.
free
session