Sourcing and bid comparison
One sourcing category, one RFP and six bids; include requirements, drafting, extraction/comparison, verification, clarification, negotiation and award approval. Exclude broader category strategy, supplier onboarding and contract migration.
Evidence: Analyst work-share scenario. Recent references remain, but none establishes these whole-workflow endpoints.
Reference task model
The reference workload starts with 100 manual staff-hours. Organization settings adjust the hours below at the same work volume. Work shares and task reductions are assumptions informed by the cited evidence. Negative values add checking work.
| Operating work | Before staff-hours | Lower saving assumption | Higher saving assumption | Staff-hours saved lower case | Staff-hours saved higher case |
|---|---|---|---|---|---|
Agree requirements and evaluation criteriaAssumption and referencesAnalyst assumption: light template/retrieval help; stakeholder decisions remain. Sources: 1 | 15 | 0% | 10% | 0 | 1.5 |
Draft and revise the RFPAssumption and referencesAnalyst transfer informed by BCG's 40% tender-creation and approximate 50% drafting cases; not an exact pooled estimate. Sources: 1 | 25 | 30% | 50% | 7.5 | 12.5 |
Extract terms and prepare a six-bid comparisonAssumption and referencesAnalyst transfer from the approximate 50% offer-comparison case; preserve readable evidence links and human interpretation. Sources: 1 | 35 | 30% | 50% | 10.5 | 17.5 |
Verify prices, omissions and competitive claims; resolve clarificationsAssumption and referencesAnalyst safeguard: no review saving; up to three extra checking hours per 100 current workflow hours. Sources: 3 | 15 | -20% | 0% | -3 | 0 |
| Negotiate and approve the award | 10 | 0% | 0% | 0 | 0 |
| Reference workload total | 100 | 15 | 31.5 |
15–31.5% before outward rounding to 15–35%. These are scenario endpoints, not confidence limits.
How much does the work mix matter?
The work shares above have not been measured. If independent checking takes 30 of the original 100 hours, instead of 15, the same task-saving assumptions produce 5–30%.
Other work shares shrink proportionately so the baseline remains 100 hours. The additional 15-hour review share is an illustrative sensitivity, not a measured allocation. An organization with a different work mix needs its own baseline.
Citations
Maximizing Value Potential from AI in 2025: Procurement
Anonymous consulting delivery cases · 2025-02
BCG reports 40% less tender-creation time at an energy provider and roughly 50% improvements in tender drafting and offer comparison at a manufacturer. The models, measurement dates and review protocol are undisclosed.
Evidence period: Case measurement and model dates undisclosed. Manufacturer took 12 months from first workshop to deployment before publication; project therefore began in 2024 or earlier, but result measurement year is unconfirmed.
Use here: Task-specific case results inform the existing analyst transfer, but do not directly measure the 15-35% complete-event range.
Primary source ↗How GATX is Scaling Procurement with an AI-First Strategy
Vendor recap of named customer webinar · 2025-07-21
GATX reports ten times the RFP capacity with the same headcount, and bid-comparison output in 10-15 seconds rather than hours. These are customer reports; the account does not time the full sourcing event or subsequent human checking.
Evidence period: Reports a recent webinar with GATX's procurement director. Deployment and measurement dates and model version undisclosed.
Use here: Capacity and subtask speed evidence; does not establish a percentage of total event labor saved.
Primary source ↗Do Humans and GAI See Eye to Eye? Implications of LLM Scoring Volatility in Supplier Evaluations
Peer-reviewed controlled supplier-scoring benchmark · 2026-05-14
A 2026 study applied three 2025 reasoning models to 123 bids. Compliance checks aligned more consistently with human judgments than competitive scoring. It supports retaining human award judgment; it does not measure labor savings.
Evidence period: AI evaluation uses 2025 reasoning models o3, Grok-3-Mini and DeepSeek R1-0528. Exact run dates unreported. Input bids came from 2023-2024 solicitations.
Use here: No time-saving measure.
Primary source ↗How to interpret this range
Conditional planning scenarios, not measured population bounds or confidence intervals.
The same work is completed manually using existing systems and approved inputs. Sources and calculation models are usable. These are workflow reference cases, not company types.
- The workflow is in steady-state use and passes checks for equivalent output quality.
- Volume, coverage and output remain unchanged. Saved capacity is not assumed to reduce staffing.
- Human review, corrections, exceptions and existing approvals are included.
- Baseline task shares and transferred task-saving rates are analyst assumptions, not measured industry allocations.
Savings include conventional extraction, matching, retrieval and integration delivered with each scoped project. The separate marginal contribution of GenAI is not established.
Separate costs: Implementation and training, Separate project/platform upkeep, Upstream remediation outside the project, Additional output and newly worthwhile work.
Already-automated work, low adoption, difficult exceptions or unreliable output can produce smaller savings, zero or additional work. The positive lower endpoint is not a guaranteed minimum.
- RFP drafting and bid comparison together represent 60% of this modeled workflow; all shares are analyst judgments.
- The tool creates a source-linked comparison, while buyers retain commercial judgment, negotiations and award decisions.
- RFP and comparison savings are bounded below the most dramatic runtime/capacity claims and around the BCG task references.
- At the low endpoint, checking AI extraction and omitted terms increases the review block by 20%.
Convert to monthly hours: Monthly operating hours saved = baseline monthly workflow hours × saving percentage. Net capacity released = operating hours saved − annual project upkeep / 12 − allocated shared upkeep / 12. Shared upkeep is allocated once across the portfolio.
What would improve the estimate
- Current shares of drafting/comparison versus negotiation and clarification
- Bid complexity and document quality
- Time needed to verify price normalization and omitted commercial terms
- Differentiated bids require extensive expert interpretation
- Unreadable files, ambiguous units or missing exclusions produce lengthy correction loops
- Existing e-sourcing templates already remove most draft/comparison effort
- Time a six-bid event using matched category complexity and output quality; include stakeholder, buyer and legal labor, plus corrections and negotiation.
Reviewed 2026-10-25 · Reference model data · Research notes