Live Production Data · 90 Days Post-Deployment
A production case study of an AI-assisted 8-stage pre-sales workflow deployed across a mid-size software services firm. Real hours tracked. Real win rates measured. No projections.
IT Services & Consulting
150–500 employees
Pre-sales, Solutions Architects, Account Managers
90 days post-deployment
Hours per proposal
14.2 hrs → 3.8 hrs
Proposal win rate
up from 24% baseline
Proposals sent per month
18 → 52 (team total)
Time-to-send
down from 5.4 days
The Challenge
Pre-sales teams at software services firms spend a disproportionate share of productive hours on proposal creation — a process that’s largely manual, inconsistently structured, and hard to scale.
The firm’s pre-sales team was operating at or below industry benchmarks. Each architect built proposals differently, estimation was done by gut feel, and client-ready documents took 5–7 days from first conversation — often losing deals to faster competitors.
Inconsistent quality and missed sections across every engagement
Informal hours estimates led to frequent under-scoping and costly rework
5–7 days from first conversation to delivery
Stakeholder maps, module breakdowns, cost models rebuilt from scratch every time
Past projects weren't referenced systematically when scoping new work
IT Industry Benchmarks (Pre-AI)
| IT Industry Benchmark | Typical Value |
|---|---|
| Avg. time to create one technical proposal | 12–18 hours |
| Time-to-send from initial client conversation | 4–7 business days |
| Proposal win rate (IT services, global avg.) | 22–28% |
| Cost per proposal (SA time + coordination) | AED 2,950–8,050 |
| Proposals a solutions architect handles / month | 4–6 |
| Rework rate due to estimation errors | 35–45% |
| Stakeholder alignment cycles per proposal | 2–4 rounds |
The Solution
The platform doesn’t replace the architect — it acts as a co-pilot. Capturing client context in Stage 1, generating solution architecture in Stage 3, producing estimation logic in Stage 6, assembling a client-ready PDF in Stage 7. All within one guided environment.
Stage 1
Paste a call transcript. AI extracts requirements, budget, timeline, tech stack, and compliance needs automatically.
Stage 2
AI identifies 3–6 distinct business roles per engagement, ensuring the proposal addresses the full buying committee.
Stage 3
AI generates solution components, module breakdowns, and tech stack recommendations for architect review.
Stage 4
Platform generates user stories, user flows, and screen specifications based on the module and requirements data.
Stage 5
Comparable past engagements are surfaced automatically, giving architects reference points for pricing and scoping.
Stage 6
Complexity factors, risk buffers, and screen counts are derived from the project’s own module and story data — not generic tables.
Stage 7
One-click PDF output: executive summary, scope, timeline, cost breakdown, and commercial terms — fully structured from workflow data.
Stage 8
Client receives a secure approval link by email. Acceptance or rejection is captured, timestamped, and logged automatically.
Paste a call transcript and the platform extracts requirements, budget, timeline, tech stack, and compliance needs automatically — the entire intake is structured in minutes.
Hours are derived from the project’s own module breakdown, story counts, and complexity factors — not gut feel or generic rules of thumb. Architects reported higher confidence from day one.
Similar past projects surface automatically during review, giving architects reference points for pricing, timeline, and risk — without searching shared drives or asking colleagues.
Results After 90 Days
The platform records total time per proposal in seconds, broken down by session — total sessions, average session length, longest session. All efficiency figures below are derived from this data.
| Metric | Before (self-reported) | After (system-tracked) | Change |
|---|---|---|---|
| Avg. hours per proposal | 14.2 hrs | 3.8 hrs | −73% |
| Time-to-send (days) | 5.4 days | 1.6 days | −70% |
| Proposals per SA per month | 4–5 | 12–15 | +3× throughput |
| Rework rate | ~40% | ~12% | −70% |
| Estimation confidence (architect-rated) | Low / Medium | Medium / High | Significant lift |
Note on time tracking: the platform records total time spent per proposal in seconds, broken down by session — total sessions, average session length, longest session. Figures above are derived from this data, aggregated across the team over the measurement period.
All required sections covered, every time. Manually created proposals achieved this only ~55% of the time.
Projects scoped through the platform came within ±15% of actual delivery hours in 7 of 10 cases. Historical error margin was ±40%.
Distinct business roles identified per engagement — proposals addressed the full buying committee, not just the technical contact.
Section structure, terminology, and financial presentation were consistent across all proposals regardless of which architect created them.
| KPI | Before | After | Change |
|---|---|---|---|
| Proposal win rate | 24% | 38% | +14 pts |
| Avg. deal size (closed proposals) | AED 679,000 | AED 881,000 | +30% |
| Time SA spent on admin vs. client work | 60% admin / 40% client | 25% admin / 75% client | Inverted |
| Proposals sent per month (team total) | 18 | 52 | +189% |
Industry Benchmarks
The platform moved every key metric from industry average to world-class or beyond. All outcome figures are from 90 days of live production use.
| Benchmark | Industry Average | World-Class Target | Platform Outcome |
|---|---|---|---|
| Hours per proposal | 12–18 hrs | < 4 hrs | 3.8 hrs ✓ |
| Time-to-send | 4–7 days | < 2 days | 1.6 days ✓ |
| Proposals per SA per month | 4–6 | 12+ | 12–15 ✓ |
| Win rate | 22–28% | 35%+ | 38% ✓ |
| Estimation error margin | ±35–45% | ±15% | ±15% ✓ |
| Rework rate | 35–45% | < 15% | ~12% ✓ |
Sources: APMP (Association of Proposal Management Professionals) 2024 benchmark report; Forrester IT Services Sales Efficiency study 2023; TechServ Alliance pre-sales productivity index 2024.
Solutions Architect Time Allocation
The platform shifts the architect’s role from content creation to content validation — a fundamentally faster mode of work. Here’s exactly how time moved.
Research & requirements gathering
3.5 hrs
Stage 1 — AI transcript analysis
Writing solution description & modules
3.5 hrs
Stage 3 — Solution review & editing
Estimation & team composition
3.0 hrs
Stage 6 — Estimation review
Costing & commercial terms
2.0 hrs
Stage 7 — Costing & terms
Document formatting & PDF assembly
1.7 hrs
Final review, PDF, sign-off setup
Coordination / overhead
0.8 hrs
Research & requirements gathering
3.5 hrs
Writing solution description & modules
3.5 hrs
Estimation & team composition
3.0 hrs
Costing & commercial terms
2.0 hrs
Document formatting & PDF assembly
1.7 hrs
Stage 1 — AI transcript analysis
0.5 hrs
Stage 3 — Solution review & editing
Stage 6 — Estimation review
Stage 7 — Costing & terms
Final review, PDF, sign-off setup
Coordination / overhead
0.8 hrs
Hours freed per proposal
SA throughput multiplier
SA time now on client work
Time freed per proposal redirected to client engagement, discovery calls, and pipeline expansion.
What Made the Difference
The efficiency gains didn’t come from working faster — they came from changing the nature of the work itself. Three shifts drove the result.
Rather than building a proposal from scratch, architects reviewed and refined AI-generated content. The cognitive load shifted from creation to validation — a fundamentally faster mode of work.
The estimation engine derives hours from the project’s own module breakdown, story counts, and complexity factors — not gut feel or generic tables. Architects reported higher confidence in their numbers from day one.
Similar past projects were surfaced during the review stage, giving architects reference points for pricing, timeline, and risk — without searching shared drives or asking colleagues.
The highest-leverage change was cognitive, not just mechanical. Reviewing a structured AI draft is significantly faster than authoring — and architects consistently applied their expertise to what actually matters: judgment calls, not document structure.
Time tracking per proposal, per stage — every session is recorded in seconds. The platform shows total hours invested by the pre-sales team per engagement. No more estimation of effort; the system simply knows.
From transcript intake through digital sign-off, everything runs inside the same environment. Eliminating the coordination overhead between tools was itself worth hours per proposal.
Beyond This Engagement
This implementation demonstrates how Carmatec turns complex, knowledge-intensive business processes into structured, production-ready AI applications.
The solution combined transcript analysis, workflow-specific AI generation, data-driven estimation, institutional knowledge retrieval, human review, and complete activity tracking within one auditable platform. Rather than introducing AI as a standalone feature, it embedded intelligence directly into the team’s existing operational workflow.
The result is not simply faster task completion. It is a more consistent, measurable, and scalable operating model in which AI strengthens human judgement and allows teams to focus on higher-value decisions.
The same approach applies to
Proposals, scoping, stakeholder mapping, deal tracking
CV scoring, shortlisting, pipeline management, onboarding
Ticket triage, resolution routing, knowledge retrieval
Automated report generation, anomaly detection, commentary
Structured extraction, review workflows, audit trails
Institutional memory retrieval, policy Q&A, decision logging
Work With Us
Carmatec works with UAE businesses to identify high-impact AI opportunities and develop secure, scalable applications aligned with their operational requirements.
Book a 30-minute AI Transformation and Product Roadmap Review with our team to explore where AI can reduce manual effort, improve decision-making, and create measurable business value across your organization.
English (UAE) English (UAE)
العربية