Estimator AI Experience Vision

Creating a five-year product vision for AI-assisted pre-construction decisions

Role: Senior UI/UX Designer
Company: STACK Construction Technologies
Domain: Construction SaaS, takeoff, estimation, pre-construction, AI-assisted workflows
Users: Estimators, subcontractors, pre-construction teams, project managers
Methods: Expert interviews, JTBD framing, competitive research, technology benchmarking, workflow mapping, Crazy 8s, Design Studio, wireframing, prototyping, leadership review
Keywords: AI product design, agentic UX, construction SaaS, pre-construction, estimator workflow, wireframe, prototype, vibe-coding, human-in-the-loop, AI-assisted decision-making, LLM, RAG, design strategy, experience vision, product vision

Overview

STACK’s Takeoff and Estimation workflow was built around manual user tasks: reading plans, measuring quantities, mapping materials, calculating costs, and preparing proposals.

As AI capabilities advanced, we saw an opportunity to rethink the product experience more fundamentally.

The question was not simply:

How can AI make takeoff faster?

The deeper question was:

How might AI help estimators decide which projects to pursue, reduce proposal risk, and bid with more confidence?

I worked with Product, Engineering, and leadership to explore a future AI-first experience for pre-construction — one where the user moves from doing every step manually to reviewing, validating, and refining AI-generated work.

What is an Experience Vision?

An Experience Vision is a future-state story of how a product could work when the company, technology, and user needs are better aligned.

It is not a feature spec.
It is not a roadmap.
It is not a fantasy UI.

It is a strategic design artifact that helps a team imagine a credible future and work backward from it.

For this project, the goal was to look beyond the next feature release and ask:

What could pre-construction estimating feel like five years from now if AI became deeply integrated into the workflow?

The vision helped the team align around a shared direction:

what customer problem we were really solving
which parts of the workflow AI could eventually support
where human judgment needed to stay in control
how the product could evolve beyond manual takeoff and estimation
what kind of experience could differentiate STACK in the market

The value of the work was not only the wireframes. The value was creating a common target.

A strong product vision gives teams something to react to, debate, refine, and build toward. It helps Product, Design, Engineering, Customer Success, and leadership move from scattered AI ideas to a shared long-term direction.

In this case, the vision helped shift the conversation from:

“How can AI make takeoff faster?”

to:

“How can AI help estimators decide what to bid, reduce risk, and create stronger proposals?”

That shift gave the company a more meaningful AI opportunity: not just automation, but better decision-making.

1. Situation

Manual workflows were becoming a strategic limitation

STACK’s existing Takeoff and Estimation workflow reflected how estimators traditionally worked: users manually reviewed plans, created measurements, mapped materials, calculated costs, and generated proposals.

This workflow was familiar, but it was also task-heavy.

At the same time, the company was preparing for broader AI adoption. A new VP Product had aligned the organization around business objectives, including AI capabilities and the preparation of a new product/SKU opportunity.

The challenge was to imagine what the product experience could become over the next 4–5 years as AI became more capable.

The product challenge

The initial focus was AI in the Takeoff module.

Takeoff is the step where estimators measure quantities directly from construction plans. But as we explored the opportunity, it became clear that takeoff was only one part of a much larger decision workflow.

Estimators were not using STACK because they loved drawing takeoffs or building estimates.

They were using STACK to:

decide whether a project was worth bidding on
understand project scope and risk
create accurate proposals faster
bid on more relevant opportunities
win more profitable work

That shifted the framing.

The real job-to-be-done was not “create faster takeoffs.”
It was “help me bid more, bid smarter, and win better projects.”

Another key insight, found in user’s early evaluation of the project.

Before estimating… Help me determine if this project is for me.”

Mapping the future workflow helped the team separate today’s manual tasks from tomorrow’s AI-assisted decision points. The exercise reframed the product around the estimator’s real goal: assess the project, manage risk, and prepare a stronger bid.

Key constraints

The vision had to work within real product and technical constraints:

the current workflow was familiar to users
the existing backend and calculations still mattered
the legacy tech stack could slow down rapid MVP development
user trust would depend on validation, not blind automation
AI capabilities were promising, but still emerging

This meant the experience needed to evolve progressively.

The goal was not to replace the estimator.
The goal was to help the estimator make better decisions, faster.

2. Actions

I helped reframe the problem around the customer’s real job

We started with a broad and ambiguous question: what should an AI-first estimation experience become?

To avoid jumping directly into features, I suggested using a Double Diamond process combined with an Experience Vision approach.

The goal was to first clarify the problem, then explore what a future experience could look like.

Together with the Product Manager and technical leads, we used:

expert interviews with internal SMEs and key customers
transcript synthesis
JTBD framing
competitive review
technology benchmarking
workflow mapping
AI capability assumptions
leadership alignment sessions

The research surfaced a critical insight:

Our customers’ business is to build things — not to draw takeoffs or estimate costs.

STACK’s value was not just helping users complete tasks faster. It was helping them make better pre-construction decisions.

We identified where AI could change the workflow

With Product and Engineering, we explored which parts of the workflow could become semi-automated or fully automated over time.

We assumed AI would progressively improve at:

understanding the user’s company, trade, and project context
reading ITBs, plans, specifications, and scope documents
identifying relevant project information
measuring quantities directly on drawings
recommending materials for takeoffs
calculating waste, labour, overhead, and profit margin
generating a bid proposal for review and submission

This helped the team imagine a different product model.

Instead of asking users to perform every step manually, the future experience could guide them through an AI-assisted workflow where the system generates results and the user validates them.

We defined the experience principles

The future experience needed to be powerful, but also understandable and trustworthy.

I shaped the concept around a few core principles:

Human-in-the-loop validation

AI could generate measurements, materials, estimates, and proposals, but users needed clear moments to review, correct, approve, or reject results.

Guided workflow

The experience should guide users step-by-step from project intake to proposal submission, reducing ambiguity without hiding the process.

Trade-specific intelligence

A concrete subcontractor, electrician, roofer, and landscaper do not think about the same project in the same way. The system needed to understand the user’s trade, context, and desired outcome.

Decision support before production work

Before users invest time in takeoff and estimation, AI should help them answer a higher-value question:

Is this project worth bidding on?

I facilitated ideation and translated the vision into wireframes

Once the team had a clearer problem definition and a future workflow, we moved into concept generation.

The Product Manager and I invited other PMs, SMEs, and designers into a short ideation session. We used Crazy 8s and a Design Studio format to generate and compare directions quickly.

I then translated the strongest ideas into wireframes for a future AI-first estimation experience.

The concept included:

status-based project navigation
an AI assistant embedded in the workflow
an Analysis step for GO / NO-GO decision support
a step-by-step review process
project document intake
AI-generated project analysis
takeoff and measurement review
material and labour estimation
proposal generation
human validation at key steps

The design intent was to make AI feel like a guided expert assistant, not a magic black box.

1. Project Analysis for GO/NO-GO

The wireframes explored how an estimator might move through an AI-assisted workflow: upload documents, review project analysis, decide GO / NO-GO, validate generated takeoffs, refine estimates, and produce a proposal.

2. Master and Specialized Agents – 6 Steps HIL

We thought it would be possible in the future to have multiple specialized agents working in the background for a master agent. Each specialized agents are trained and RAG specific documents and skills relevant to their tasks: understanding complex plan sets, counting, measuring and sourcing materials, adding labour, waste, profits, etc. and writing a solid bid proposal.

3. Human in the Loop

Each specialized agents are working in the background in a semi-linear flow. User must review and validate each steps in a specific order, for agents to use decisions made with the previous agent. Each steps have a different toolset – graphic tools to edit takeoffs on a plan, recommender-system to change or source materials elsewhere, spreadsheet and cost editor, etc. And… the user can use the LLM to change, adjust at each steps.

We explored agentic workflows and trade-specific assistants

The vision also explored how multiple AI agents could support the workflow.

At a high level, the model included:

a trade-specific agent that understands the user’s specialty and context
expert agents for different workflow steps
document processing and computer vision capabilities
RAG and system prompts to ground the assistant in project documents
a future orchestrator to coordinate specialized agents

The proposed workflow imagined agents supporting:

Project analysis → Takeoffs → Materials → Estimation → Proposal generation

But the user would remain in control through validation checkpoints.

This was important. In construction, accuracy and trust matter. A wrong quantity, scope assumption, or proposal detail can create real financial risk.

So the experience had to help users understand what the AI found, where it came from, and what needed review.

We reviewed the concept with leadership and product teams

The wireframes were presented to leadership and product managers to gather feedback across three lenses:

Viability

The value of a human-in-the-loop AI workflow was well received. If STACK could help users assess project fit and proposal risk earlier, it could create a meaningful competitive advantage.

Feasibility

The technical feasibility was still emerging, but not unrealistic. Based on the direction of LLMs, RAG, agent orchestration, and document processing, the concept helped the organization reason about what could become possible over time.

Desirability

Although user adoption could not be fully validated from wireframes alone, Customer Success saw the concept as a promising improvement over a purely manual-step application.

We simplified the experience after feedback

The first concept generated useful discussion, but we also saw that some people had difficulty understanding parts of the flow and interface.

With feedback from the presentation — and using AI tools such as ChatGPT and Figma Make to challenge our own assumptions — we simplified the design direction.

Key improvements included:

adding a clear GO / NO-GO decision point
simplifying and empowering the navigation menu
condensing the flow from 7 to 5 clearly labeled steps
clarifying where each project was in the workflow (new, active, bidding, archive…)
making the step-by-step process more visible with numbers 1-5

This iteration made the experience easier to understand and more directly connected to the estimator’s decision-making process.

Polished Prototype

Major improvements were made to the navigation and menu features. The navigation is streamlined to the GO / NO-GO and the 5 steps. While the documents are being processed, we ask the user some questions about the setup he had in mind for this job (trade, crew, library) to have better, in-context analysis. We removed the “Documents” step and merged it with the Analysis step; this is where user review the files and AI recommendations before creating takeoffs or estimates, the assistant helps the estimator understand project fit, scope, risk, and whether the opportunity is worth pursuing.

3. Results

A clear AI-first product vision for pre-construction

The work produced a future-state experience vision for an AI-assisted estimation product.

The vision reframed the product from a manual workflow tool into a guided decision system for pre-construction teams.

It showed how AI could help users:

understand project documents faster
assess project fit and risk
decide whether to bid
generate and validate takeoffs
map materials and labour assumptions
estimate overhead, profit, and alternates
generate proposal content
review and approve AI-generated outputs

The result was not a production feature launch. It was a strategic product vision and prototype direction that helped the organization understand what an AI-first STACK experience could become.

A stronger product framing

The project helped shift the internal conversation from:

“How can AI make takeoff faster?”

to:

“How can AI help estimators bid more, win more, and reduce project risk?”

That shift mattered because it connected AI investment to customer value and business outcomes.

It also gave Product, Design, Engineering, Customer Success, and leadership a shared language for discussing future AI capabilities.

A validated direction for future exploration

The concept was reviewed with leadership and product stakeholders and helped validate several strategic directions:

AI should assist the full pre-construction workflow, not only one task
human validation is essential for trust
GO / NO-GO analysis could become a high-value differentiator
trade-specific intelligence matters
agentic workflows could support a phased product roadmap
the user experience must make AI output reviewable and explainable

The work also influenced near-term thinking around MVP opportunities, including early “chat with plans” experiences and document-understanding workflows.

What this project proves

This project shows my ability to design at the intersection of product strategy, AI capability, and complex expert workflows.

I helped take an ambiguous AI mandate and turn it into a clearer experience vision: one grounded in customer jobs, workflow reality, technical constraints, and future product opportunity.

The design challenge was not simply to add AI to an existing interface.

It was:

How do we help estimators trust AI enough to make faster, better, higher-stakes business decisions?

Relevant strengths

AI product strategy
Reframing AI from feature automation to decision support and workflow transformation.

Agentic UX and human-in-the-loop design
Designing AI-assisted flows where users review, validate, correct, and approve generated results.

Complex B2B SaaS
Working inside a construction product with dense workflows, expert users, business risk, and legacy constraints.

Product discovery and facilitation
Using interviews, JTBD, competitive research, technology benchmarking, Crazy 8s, and Design Studio methods to align teams.

Systems thinking
Connecting project intake, document analysis, takeoff, materials, estimation, and proposal generation into one coherent product vision.

Strategic storytelling
Turning a future-state concept into a narrative leadership could evaluate, challenge, and build upon.

Home

More Projects

Top of Page