CONTRACT·AI
/ SCM / Category Intelligence / Use Case Roadmap v2
O&G SCM INTERNAL BUILD GENIE + AZURE 3–6 MO
AI Strategy — Contracting Domain

CONTRACT
INTELLIGENCE

A reframed use case portfolio anchored in the 20 questions every category manager should be able to answer about their contracting portfolio — but currently can't. Spend leakage management is handled separately. This brief focuses exclusively on managing the contract itself: planning, execution, risk, and commercial intelligence.

20
CM questions answered
10
Use cases
15–30%
Category efficiency gain
75%
Faster sourcing prep
Scope boundary: Off-contract spend detection and invoice price deviation are excluded — those are being addressed by the parallel spend leakage initiative. This roadmap covers everything before, during, and after the contract itself.
The 20 Questions
These are the questions category managers ask daily — or should. Today most take hours or days to answer. Every use case in this brief answers one or more of them.
ACP & Pipeline Planning
  • Q1Which of my contracts expire this year and when must I start a sourcing event to meet the deadline?UC-01
  • Q2How many sourcing events do I have planned this year and do I realistically have capacity to deliver them?UC-02
  • Q3Which contracts can be extended vs. must go through full re-competition based on policy, risk, or value?UC-01
  • Q4Which contracts have been sole-sourced 3+ years without a competitive event — and why?UC-06
  • Q5What is the critical path timeline for my most strategically important upcoming renewal this cycle?UC-02
Commercial Intelligence
  • Q6What rates and terms did I achieve last time, and how do they compare to current market conditions?UC-08
  • Q7Which contracts have escalation clauses triggering soon and what is the projected cost impact on my budget?UC-04
  • Q8Are there volume discount tiers or rebate thresholds in my contracts that we're failing to capture?UC-07
  • Q9What savings have I committed to deliver this year and am I tracking against that target by category?UC-05
  • Q10Should I renegotiate this contract early given market movement, or is it better to wait until expiry?UC-04
Supplier & Risk
  • Q11Which of my contracts represent single-source supply risk and do I have a documented contingency?UC-06
  • Q12How concentrated is my spend within this category — am I over-reliant on one or two suppliers?UC-06
  • Q13Are there financial health or external risk signals on any of my key contract suppliers right now?UC-06
  • Q14What happened the last time we went to market in this category — who bid, who won, and at what price?UC-08
  • Q15Which suppliers have a pattern of non-compliance or performance issues in contracts like this one?UC-07
Contract Health
  • Q16Are all obligations in my active contracts assigned to named owners, and are upcoming milestones being tracked?UC-07
  • Q17What non-standard or risky terms exist in my portfolio that I should target for renegotiation at next renewal?UC-09
  • Q18Which contracts have terms, entitlements, or protections that were negotiated but have never been activated?UC-07
  • Q19What do I need to know before I start drafting the next RFP or SOW for this category?UC-10
  • Q20Which upcoming renewals need to be escalated to legal or senior leadership for early engagement?UC-01
Use Cases
10 use cases mapped to the 20 questions. Click any card to expand data requirements, tech stack, delivery timeline, and Genie query examples.
00
Contract Data Extraction & Lake Foundation
GEP SMART → structured Delta tables → Genie-ready data. The enabler for all 10 use cases.
Foundation ● Medium effort
▾
Enables all Q1–Q20
Why this comes first
GEP SMART holds contract data in a mix of structured header records and PDF documents. The quality of every downstream use case depends on how cleanly this data lands in the lake. Key attributes that must be machine-readable: expiration dates, renewal flags, notice periods, contract value, SLAs, escalation clauses, obligation milestones, and rate schedules. Start with the most structured fields (dates, values, categories) and progressively tackle the complex clause extraction.
Extraction targets — priority order
1st pass (structured): Supplier, category, start/end, contract value, status, auto-renewal flag, notice period

2nd pass (LLM extraction): Escalation terms, SLAs, obligation milestones, payment terms, liability caps

3rd pass (complex): Rate card tables, penalty structures, volume tier schedules
Delivery timeline
6–8 weeks for initial pipeline. Continuous improvement ongoing.
Architecture pattern
Bronze layer: Raw GEP exports + PDF blobs → Azure Blob
Silver layer: LLM-extracted structured attributes → Delta table
Gold layer: Curated, joined, Genie-ready analytics tables

Accuracy target: 85–92% automated extraction accuracy. Human-in-loop review for complex clause types during initial ramp. Spot-check 10% of dates and values before declaring gold-layer production-ready.
Tech stack
Azure Document Intelligence Azure OpenAI GPT-4o Databricks AutoLoader Delta Lake Bronze/Silver/Gold GEP SMART API / bulk export LangChain extraction chains
Who owns this
Data engineering team + one procurement SME who validates extraction quality. Critical: a procurement person who knows what "correct" looks like must be involved from day one, not just at the end.
01
Annual Contracting Plan (ACP) Intelligence
A live, AI-enriched view of the ACP — what's expiring, when, decision status, escalation flags, and sourcing readiness by category
Quick Win ● Low effort
▾
Q1 · Q3 · Q5 · Q20
Business Value
The Annual Contracting Plan is typically a spreadsheet-maintained artifact that's out of date within weeks. This use case makes the ACP a live, self-updating intelligence dashboard. Category managers see their upcoming contract events on a calendar view with auto-calculated lead times, decision flags (extend / re-compete / sole source), and escalation signals — all driven by the data already in GEP. Leadership gets a portfolio-level ACP summary without chasing status updates.
What AI adds
Beyond the calendar, the LLM layer automatically flags: which contracts are eligible for extension based on existing terms, which require legal/leadership approval based on value thresholds, and which have conflicting deadlines creating a capacity risk. It can auto-draft a renewal decision brief — a 1-page summary of the contract, utilization, and recommended action — for each upcoming event.
Delivery timeline
3–5 weeks after UC-00 gold tables are available.
Tech stack
Databricks Genie Space Unity Catalog (contract gold table) Azure OpenAI (renewal brief generation) MS Teams / Copilot embed
Key outputs
· ACP calendar by category manager and category
· Days-to-expiry with required start date for sourcing
· Extend / re-compete / escalate recommendation flags
· Portfolio-level ACP summary for GM/Director reviews
· Auto-generated renewal decision brief per contract
Who uses it
Category Managers Procurement Directors SCM Leadership Legal
Example Genie queries (Databricks Genie Space)
Which of my contracts expire in Q3 and haven't been actioned yet?
Show all contracts where I've missed the required sourcing start date
List contracts flagged for leadership escalation this quarter
Which contracts are eligible for extension under their current terms?
Generate the ACP summary for my Drilling Services category
Which contracts expire in the next 180 days with value over $5M?
02
Sourcing Event Workload & Capacity Planner
Visualize the entire sourcing event pipeline, flag capacity crunches, and model what's achievable within the team's bandwidth
Quick Win ● Low effort
▾
Q2 · Q5
Business Value
Category managers and their teams are chronically stretched. Without a system-driven view of the full sourcing event pipeline, they discover capacity crunches after the deadline has been missed. This tool maps every planned sourcing event against realistic lead times and team capacity — showing where conflicts exist months in advance. It makes the invisible workload visible, supporting resource allocation and timeline negotiations with stakeholders.
AI enrichment
AI calculates estimated event complexity (RFQ vs. full RFP vs. strategic sourcing) based on contract type, value, and category, then assigns expected lead time ranges. It flags sourcing crunch zones where multiple high-complexity events overlap, and suggests sequencing adjustments or extension candidates to relieve pressure.
Delivery timeline
4–5 weeks. Builds directly on UC-01 ACP data.
Tech stack
Databricks Genie Space Delta Lake (ACP + team structure) Azure OpenAI (complexity scoring) Power BI / Genie visual embed
Key outputs
· Gantt-style sourcing calendar per CM and team
· Capacity heatmap by month and quarter
· Conflict alerts: overlapping events with same lead time
· Recommended sourcing event sequencing
· "What if I extend X?" scenario modeler
Who uses it
Category Managers SCM Directors Resource Planning
Example Genie queries
How many sourcing events do I have planned in Q2 and Q3 combined?
Which months in the next 12 are overloaded with sourcing deadlines?
If I extend Contract X by 6 months, how does that relieve my Q4 crunch?
Which sourcing events can be run concurrently without risk?
03
Contract Q&A — Natural Language Document Search
"What's our liability cap with Halliburton?" answered in seconds with citation. The highest daily-use contracting AI tool.
Strategic ● Med–High effort
▾
Q6 · Q14 · Q17 · Q18 · Q19
Business Value
Manual contract review averages 92 minutes per contract. Across hundreds of O&G service agreements, practitioners spend days per week just finding information that's already in their contracts. A RAG-based interface where any team member can ask a natural language question and get an answer with a clause-level citation completely eliminates this friction. Field ops asking "do we have a price revision clause with our chemical supplier?" get an answer in 8 seconds. This is the use case that changes behavior permanently.
Critical design requirements
Access control at the document level is non-negotiable. A field ops buyer must not see confidential M&A or sensitive commercial agreements. Use Azure AD groups + row-level security in the vector store. Every response must cite the contract name, date, and clause section — never a "hallucinated" answer without source attribution.
Delivery timeline
10–14 weeks for MVP. RAG pipeline + access control + Teams/Copilot interface.
Tech stack
Azure OpenAI GPT-4o Azure AI Search (Vector index) Databricks Vector Search LangChain / LlamaIndex (RAG) Copilot Studio / Teams bot Azure Blob (document storage) Azure AD (document-level RBAC)
What people actually ask
Pre-sourcing research, clause comparison across contracts, obligation discovery, term hunting for negotiations, and compliance checks before major decisions.
Example RAG queries — real practitioner questions
What is the liability cap in our current MSA with SLB?
Do any of my contracts include a most-favored-nation clause?
Summarize the payment terms across my top 10 suppliers
What are our termination rights in the pipeline logistics contract?
Which contracts have indemnity language favoring the contractor?
What did we agree to on IP ownership in the last tech services contract?
Does Contract X allow us to audit the supplier's invoices?
Which contracts have change-of-control provisions?
04
Market Timing & Price Escalation Intelligence
Combines commodity index feeds with escalation clause tracking to answer: "Should I renegotiate now, or wait?" — and forecast budget impact
Core Build ● Medium effort
▾
Q7 · Q9 · Q10
Business Value — two capabilities in one
Escalation Tracker: Long-term O&G service contracts (drilling, compression, transportation) routinely contain CPI/PPI escalation clauses. A 3% escalation on a $50M contract that wasn't forecasted is a $1.5M budget surprise. AI extracts escalation terms from PDFs, connects them to live index feeds, and proactively surfaces upcoming adjustments before the invoice arrives.

Market Timing Signal: AI compares current market rates (field services benchmarks, commodity indices) to contracted rates to flag when the window is favorable to renegotiate early — or when waiting for expiry is the better move.
Delivery timeline
8–10 weeks. LLM clause extraction + external index API integration + Genie dashboard.
Tech stack
Azure OpenAI GPT-4o (clause extraction) Azure Document Intelligence Databricks Genie BLS API / IHS Markit / commodity feeds Budget system integration (SAP/Oracle)
O&G-specific value
Oil field services cost indices move meaningfully with oil price cycles. This tool tells category managers when the market is soft enough to push for rate reductions before a contract naturally expires — capturing savings that would otherwise be left until the next cycle.
Example Genie queries
Which contracts have escalation clauses triggering in the next two quarters?
What is the projected total cost increase from escalations next fiscal year?
Show all PPI-linked contracts and their base index dates
For which categories is the current market below our contracted rate?
Which contracts are worth pursuing early renegotiation given index movement?
05
Savings Commitment Tracker
Track committed savings targets by category manager and sourcing event — from negotiation through to realized value — in real time
Quick Win ● Low effort
▾
Q9
Business Value
Procurement organizations commit savings targets at the start of the year by category and sourcing event. Tracking actual delivery against those commitments is almost always done manually in spreadsheets — updated quarterly, often too late to course-correct. This Genie space makes savings tracking continuous. Category managers can see in real time whether their negotiated savings are on track to be realized, and leadership gets a live portfolio view without the monthly status meeting.
The gap AI fills
Most savings trackers capture the negotiated savings but lose track of realized savings once spend flows. AI connects the negotiated rate change to actual invoiced spend to compute what's been truly captured vs. what remains pipeline.
Delivery timeline
3–4 weeks if savings commitment data exists in GEP or a defined source.
Tech stack
Databricks Genie Space Delta Lake (savings + ACP data) GEP SMART savings tracking export Copilot / Teams summary embed
Key outputs
· Savings committed vs. realized by category / CM / quarter
· At-risk savings events (contract delayed, scope changed)
· Pipeline of uncommitted savings opportunities
· YTD tracking vs. annual target with trend
Who uses it
Category Managers Procurement Leadership Finance / CFO office
Example Genie queries
What savings have I committed this year vs. what's been realized so far?
Which sourcing events have savings at risk due to timeline slippage?
What is the total savings pipeline for Q3 and Q4 across all categories?
Show me categories where realized savings are tracking below 80% of target
06
Supplier Concentration & Single-Source Risk Monitor
Map supplier concentration risk by category, flag sole-source contracts with no competitive history, and surface financial health signals
Core Build ● Medium effort
▾
Q4 · Q11 · Q12 · Q13
Business Value
In O&G, supplier concentration risk is operationally critical — not just a procurement governance concern. A key well services supplier going into financial distress, or a single-source chemicals supplier facing a supply shock, can halt operations. Most category managers don't have a systemic view of how concentrated their category is until something goes wrong. This use case creates a continuous, quantified risk picture by category, flagging sole-source dependencies, multi-year uncontested contracts, and external risk signals automatically.
Sole-source intelligence
AI surfaces every contract that: (1) has been sole-sourced for 3+ years, (2) has never been competitively bid, or (3) requires a justification update. It also flags which of these have upcoming renewals — so the ACP decision can include a competitive challenge where appropriate.
Delivery timeline
7–10 weeks. Contract data + optional external risk API integration.
Tech stack
Databricks Genie Space Delta Lake (contract + sourcing history) External risk API (D&B / Riskmethods) Azure OpenAI (risk narrative generation) Power BI embed (risk heatmap)
Key outputs
· Supplier concentration index (HHI) by category
· Sole-source contract register with competitive event history
· Supplier financial health flag by contract
· Risk-ranked renewal list for ACP planning
· AI-generated risk narrative per flagged supplier
Example Genie queries
Which of my categories have >70% of spend with a single supplier?
Show me all sole-source contracts that haven't had a competitive event in 3+ years
Are any of my high-value sole-source contracts up for renewal this year?
Which suppliers in my portfolio have flagged financial distress signals?
What is the single-source risk exposure in drilling services?
07
Obligation, SLA & Contract Utilization Monitor
Track what was promised (obligations, SLAs, milestones), catch unused entitlements, and claim value that's already been negotiated but never activated
Core Build ● Med effort
▾
Q8 · Q15 · Q16 · Q18
Business Value
Most O&G procurement teams negotiate good contracts, then fail to extract the full value from them. Three common leakage points: (1) SLA penalties never claimed because breach isn't tracked, (2) volume rebates or discounts not triggered because no one monitors threshold achievement, and (3) review rights and audit clauses that expire unused. This tool continuously monitors all three, extracting value from contracts that are already in place without requiring any new negotiations.
The underused entitlements problem
AI scans contracts for rights and entitlements that have a time-bound activation window — price review triggers, benchmarking rights, demand forecasting obligations — and alerts the CM before they lapse. This is pure found value from existing contracts.
Delivery timeline
8–12 weeks. Complexity depends on how much operational data (delivery logs, invoices) is available to validate SLA performance.
Tech stack
Azure OpenAI (obligation extraction) Databricks Genie dashboard Delta Lake (contract + ops data) ERP / CMMS data integration Copilot alert notifications
Key outputs
· Obligation register with owner assignment and status
· SLA breach log with penalty credit calculation
· Volume rebate attainment tracker
· Expiring entitlements alert (price review windows, audit rights)
· Supplier performance scorecard by contract
Example Genie queries
Which suppliers have breached SLA commitments this quarter?
What penalty credits are we eligible to claim from active contracts?
Which contracts have volume rebate thresholds we're close to hitting?
Show entitlement windows expiring in the next 60 days across my portfolio
Which supplier performance issues are approaching renegotiation leverage?
08
Sourcing Intelligence — Historical Contract Benchmarking
During pre-sourcing, surface comparable historical contracts, achieved rates, supplier patterns, and negotiation outcomes as structured intelligence
Strategic ● Med–High effort
▾
Q6 · Q14
Business Value
When a category manager prepares for a sourcing event, they need institutional memory: what did similar contracts achieve historically, who bid last time, what were the differentiating factors, and what concession patterns emerged in negotiations? In most O&G companies, this intelligence exists only in people's heads — and when those people change roles, the knowledge walks out the door. AI systematically captures and surfaces this for every future sourcing event.
Auto-generated sourcing brief
For each new sourcing event, the system auto-generates a pre-negotiation brief: comparable contracts with achieved rates, which suppliers bid and how they compared, what terms were most contested, recommended walk-away thresholds, and market context. A task that previously took a CM 2–3 days of digging now takes 90 seconds.
Delivery timeline
12–16 weeks. Requires UC-03 (Contract Q&A) as foundation for deep term access.
Tech stack
Azure OpenAI (brief generation) Databricks Genie + Vector Search Delta Lake (contract history) Embeddings-based similarity matching GEP SMART sourcing event integration
Note
This use case gets dramatically better the more contract history exists in the lake. Start capturing structured sourcing outcomes now — even from GEP SMART — so the model has data to learn from. Value improves linearly with data depth.
Example Genie queries during pre-sourcing research
What day rates did we achieve for drilling services in the last 3 years?
Who were the bidders in our last completion services sourcing event?
What payment terms did we negotiate with Supplier X historically?
Which suppliers have consistently accepted volume discount tiers above 12%?
Generate a pre-negotiation brief for our upcoming well services renewal
09
Clause Risk Scoring & Portfolio Risk Audit
AI pre-screens incoming supplier paper against your playbook AND audits your existing portfolio for non-standard risk exposure
Strategic ● Med–High effort
▾
Q17 · Q20
Business Value
In O&G, contractor agreements carry some of the most contentious indemnity and liability language in any industry. Non-standard terms that accumulated in contracts over years — uncapped liability, unusual IP provisions, waiver of consequential damages — become visible only when a dispute arises. This use case does two things: pre-screens new supplier paper before legal review, AND audits the existing portfolio to create a prioritized list of terms to renegotiate at next renewal opportunity. The portfolio audit alone often surfaces millions in latent risk exposure.
What makes this O&G specific
The CUSA/COUSINS indemnity doctrine, knock-for-knock provisions, pollution liability, and force majeure language in energy contracts require domain-specific playbook calibration. This is a joint legal + procurement build.
Delivery timeline
12–16 weeks. The legal playbook codification is the long pole in the tent.
Tech stack
Azure OpenAI GPT-4o Azure Document Intelligence Clause classification fine-tuning Databricks (clause analytics) Power Apps (legal review workflow)
Risk output tiers
🔴 High: Uncapped liability, one-sided IP transfer, broad indemnity
🟡 Medium: Non-standard payment, escalation without cap
🟢 Acceptable: Within playbook tolerance
⚪ Missing: Required clauses absent
Example Genie queries
Which active contracts have high-risk clause deviations flagged?
Summarize non-standard terms in my portfolio that should be targeted at next renewal
Which contracts have missing termination-for-convenience clauses?
Show contracts where our liability exposure is uncapped
10
AI-Assisted Sourcing Document Drafting
Generate first-draft RFPs, SOWs, evaluation criteria, and supplier scorecards from historical contracts + category context — eliminating the blank page
Core Build ● Medium effort
▾
Q19
Business Value
RFP and SOW drafting is one of procurement's most time-consuming tasks. Category managers spend days creating documents that are structurally similar to what was done last time, but with updated scope and terms. AI reduces this from 2–3 days to 2–3 hours by generating a first draft that incorporates: the prior contract's scope language, terms learned from the existing agreement, evaluation criteria from comparable past events, and standard clause libraries. The CM reviews, adjusts, and approves — the blank page problem disappears.
What the AI knows
Sourcing context is richer when UC-03 (Contract Q&A) and UC-08 (Sourcing Intelligence) are operational — the drafting tool pulls directly from prior contract terms, historical supplier performance, and category benchmarks to populate relevant sections automatically.
Delivery timeline
8–10 weeks. Drafting tool + template library + Copilot Studio integration.
Tech stack
Azure OpenAI GPT-4o Copilot Studio (Word / Teams) Databricks (contract context retrieval) RAG over prior contracts + clause library GEP SMART integration (publish output)
Documents AI can draft
· RFP / RFQ / RFI
· Statement of Work (SOW)
· Evaluation criteria and scoring matrix
· Supplier pre-qualification questionnaire
· Renewal notice letter to supplier
· Internal approval memo
Example AI drafting prompts
Draft an RFP for well services renewal based on the prior MSA with SLB
Generate evaluation criteria for a chemicals supply sourcing event
Create a supplier scorecard template for our upcoming drilling services bid
Draft the scope of work section using our last pipeline services contract as reference
Recommended Build Sequence
Ordered by value delivery speed, respecting the dependency chain. UC-00 unlocks everything.
SeqUse CaseTierEffortDeliveryAnswersKey Dependency
1 00 · Contract Data Extraction Pipeline Foundation Medium Wk 1–8 Q1–Q20 (enables all) GEP export access + Azure Blob
2 01 · ACP Intelligence Dashboard Quick Win Low Wk 4–8 Q1, Q3, Q5, Q20 UC-00 gold tables live
3 05 · Savings Commitment Tracker Quick Win Low Wk 5–8 Q9 Savings data in GEP or defined source
4 02 · Sourcing Workload & Capacity Planner Quick Win Low Wk 6–10 Q2, Q5 UC-01 ACP data
5 04 · Market Timing & Escalation Intelligence Core Build Medium Wk 8–12 Q7, Q9, Q10 LLM clause extraction + index API feeds
6 06 · Supplier Concentration & Single-Source Risk Core Build Medium Wk 8–12 Q4, Q11, Q12, Q13 UC-00 + optional external risk API
7 07 · Obligation, SLA & Utilization Monitor Core Build Medium Wk 10–14 Q8, Q15, Q16, Q18 Ops data (ERP/CMMS) integration
8 10 · AI Sourcing Document Drafting Core Build Medium Wk 10–14 Q19 Template library + Copilot Studio
9 03 · Contract Q&A (RAG) Strategic Med–High Wk 12–18 Q6, Q14, Q17, Q18, Q19 Document storage + vector pipeline + RBAC
10 08 · Sourcing Intelligence / Benchmarking Strategic Med–High Wk 14–20 Q6, Q14 UC-03 (RAG) operational + history depth
11 09 · Clause Risk Scoring & Portfolio Audit Strategic Med–High Wk 14–22 Q17, Q20 Legal playbook codification (legal engagement)
Data & Architecture Flow
From GEP SMART source data through to practitioner-facing Genie interaction
01 · SOURCE
GEP SMART
  • Contract records (structured)
  • PDF documents
  • Sourcing event history
  • Savings tracking
  • Supplier data
02 · INGEST
Raw Data Lake
  • Azure Blob (PDFs)
  • Databricks AutoLoader
  • Delta Lake Bronze
  • ERP / CMMS feeds
  • External index APIs
03 · ENRICH
AI Processing
  • Azure Doc Intelligence
  • GPT-4o extraction
  • LangChain chains
  • Vector embedding
  • Clause classification
04 · SERVE
Analytics Layer
  • Delta Lake Gold tables
  • Unity Catalog (RBAC)
  • Databricks Vector Search
  • SQL Warehouse
  • MLflow models
05 · USE
Practitioner Interface
  • Databricks Genie Spaces
  • MS Teams Genie embed
  • Copilot Studio bots
  • Power Apps workflows
  • Word/GEP integration
How to Frame the Team Conversation
Start with the 20 questions as the opening provocation — ask the team which ones they currently cannot answer in under 5 minutes. That gap IS the opportunity. Most O&G procurement teams will struggle with 14+ of them, which immediately establishes the business case without needing ROI modeling upfront.

First sprint recommendation: UC-00 (extraction pipeline) + UC-01 (ACP dashboard) + UC-05 (savings tracker) delivered together. These three use cases solve the most frequently voiced pain points — "I don't know what's expiring," "I can't track our savings commitments," and "the ACP spreadsheet is always wrong" — and prove the Databricks Genie platform works before committing to the heavier builds.

The strategic anchor: UC-03 (Contract Q&A) should be the signature use case that shows what this program is ultimately capable of. Position it as the "12-week milestone" — the moment when any practitioner can ask any question about any contract and get an answer with a citation. That demonstration changes the conversation from "AI project" to "this is how we work now."