Recurring workflows that AI-native operations can take on.
A library of 31 workflows, grouped by business function. For each one: what an AI-native version looks like, what stays with people, how deep the AI change typically goes, and the signal that it is worth a closer look. Use it to pick the one workflow you want examined first.
A starting menu, not a promise.
Every row is a pattern we would examine. It is not a result we claim for your business. Whether a workflow is worth changing depends on its volume, its exceptions and its controls, and only your evidence can show that. The free diagnostic gives a first read on one workflow. The Operational Leverage Audit measures it.
The "Typical AI depth" column uses the same four levels as the free diagnostic's Operational Leverage Snapshot:
| Level | What it means | What changes |
|---|---|---|
| Level 1: Task automation | Individual steps are automated. The workflow itself stays the same. | Less manual keying, copying and checking. People still run the process. |
| Level 2: Workflow augmentation | AI prepares, drafts or recommends inside the existing workflow. A person decides. | Faster preparation and fewer touches. Decisions and sign-off stay human. |
| Level 3: Workflow redesign | The workflow is rebuilt around agents, deterministic checks and human exception handling. | Routine cases run end to end. People handle exceptions and set the rules. |
| Level 4: Re-architecture | The operating model around the workflow changes: systems, roles and controls are redesigned together. | Capacity grows without matching headcount. Governance is built into the system, not added afterwards. |
Depth and fit signals are typical, not fixed. The same workflow can sit at Level 1 in one business and Level 3 in another. Every row is illustrative unless it points to one of our own AdTech case studies.
Getting from interest to signed revenue with fewer hand-offs.
Sales operations carry a lot of coordination: routing, record-keeping, quoting and chasing. Most of it follows rules. See also SaaS & technology and professional services.
| Workflow | What AI-native looks like | What stays human | Typical AI depth | Typical fit signal |
|---|---|---|---|---|
| Inbound lead qualification and routing | Each inbound lead is enriched, scored against your stated criteria and routed to the right owner, with the reasoning logged. | Defining what a good lead is. Handling strategic or unusual accounts. | Level 2 | High volume, clear criteria |
| CRM data hygiene and enrichment | Duplicates, missing fields and stale records are flagged and fixed against defined rules before they spread into reports. | Merges that affect ownership or commission. Rule changes. | Level 1 | Repetitive, rule-based, low risk |
| Proposal and quote preparation | A draft proposal is assembled from approved templates, price lists and the deal record, then checked against discount rules. | Pricing exceptions, negotiation and final sign-off. | Level 2 | Templated output, repeatable inputs |
| Renewal and churn-risk monitoring | Usage, support and billing signals are watched continuously. At-risk accounts surface with the evidence attached. | The customer conversation and any commercial concession. | Level 2 | Signals already sit in your systems |
| Sales commission calculation | Commissions are calculated from closed deals and plan rules, with every line traceable to its source record. | Plan design, disputes and approvals. | Level 1 | Clear rules, costly disputes |
Routine cases resolved, hard cases routed to the right person.
Support work is high volume and policy-bound, with a long tail of exceptions. That shape suits agents with verification and a clean hand-off to people. See also e-commerce & marketplaces and SaaS & technology.
| Workflow | What AI-native looks like | What stays human | Typical AI depth | Typical fit signal |
|---|---|---|---|---|
| Ticket triage and routing | Incoming tickets are classified, tagged by urgency and routed to the right queue, with low-confidence cases flagged for a person. | Queue design and anything flagged as sensitive. | Level 1 | High volume, clear categories |
| First-response drafting | A draft reply is prepared from your knowledge base and the customer's history, with sources shown to the agent who sends it. | Tone, judgement and the decision to send. | Level 2 | Repeated questions, good documentation |
| Refund and return eligibility | Each request is checked against policy, order data and fraud rules. Clear cases are settled; the rest go to a person with the evidence. | Exceptions, goodwill gestures and policy changes. | Level 3 | Written policy, many similar requests |
| Customer onboarding and account setup | Setup steps across systems run as one tracked workflow, with checks before each account goes live. | Non-standard contracts and enterprise configurations. | Level 3 | Many systems, repeated hand-offs |
| Complaint and escalation handling | Complaints are summarised, linked to the order and contract history, and prioritised for the owner. | The response, the remedy and any admission of fault. | Level 2 | High stakes, slow information gathering |
Matching, chasing and reconciling, checked before it commits.
Back-office finance is mostly coordination plus exceptions. Money is involved, so verification gates matter more here than anywhere. See also finance operations.
| Workflow | What AI-native looks like | What stays human | Typical AI depth | Typical fit signal |
|---|---|---|---|---|
| Invoice intake and three-way match | Invoices are read, matched to purchase orders and receipts, and queued for payment only when every check passes. | Mismatches above tolerance and payment release. | Level 3 | High volume, clear matching rules |
| Accounts receivable collections | Reminders, invoice questions and payment-plan requests are handled in sequence, with escalation rules set by you. | Disputes, write-offs and key-account relationships. | Level 3 | Many small balances, repeated follow-up |
| Expense review against policy | Each claim is checked against policy, receipts and limits. Clean claims pass; outliers are explained and flagged. | Policy exceptions and approval of flagged claims. | Level 2 | Written policy, high claim count |
| Month-end reconciliation | Accounts are reconciled continuously, not at month end. Breaks are traced to source and explained before close. | Judgemental adjustments and sign-off of the close. | Level 3 | Recurring breaks, spreadsheet-heavy close |
| Vendor onboarding and document checks | Supplier documents are collected, checked for completeness and validity, and compared with existing records. | Risk decisions and approval of new suppliers. | Level 2 | Checklist-driven, document-heavy |
Keeping orders moving when something goes wrong.
The happy path is often automated already. The cost sits in the exceptions: holds, delays, claims and bad data. See also e-commerce & marketplaces.
| Workflow | What AI-native looks like | What stays human | Typical AI depth | Typical fit signal |
|---|---|---|---|---|
| Order exception handling | Address, stock and payment holds are diagnosed and resolved within set rules, with the customer kept informed. | Holds that need a commercial or fraud decision. | Level 3 | Frequent, similar exceptions |
| Inventory replenishment proposals | Reorder proposals are drafted from sales velocity, lead times and stock levels, with assumptions shown. | Purchase approval and supplier strategy. | Level 2 | Many SKUs, stable demand data |
| Supplier confirmation and ETA chasing | Purchase orders are confirmed, delivery dates are tracked and late suppliers are chased automatically. | Renegotiation and supplier escalation. | Level 2 | Repetitive email follow-up |
| Logistics claims | Lost and damaged shipment claims are assembled with evidence, filed with carriers and tracked to payment. | Disputed claims and carrier relationships. | Level 3 | Evidence-heavy, deadline-driven |
| Product and catalogue data onboarding | Supplier product data is normalised, validated against your schema and checked before it goes live. | Brand, pricing and content standards. | Level 2 | Messy inputs, fixed target format |
Where our own production work sits.
This is the one function where we can point to our own systems: header bidding, an AdCP proof-of-concept, an analytics copilot and IVT detection, described in Case Studies. Rows marked as our work link there. The others are patterns we would examine. See also AdTech & programmatic.
| Workflow | What AI-native looks like | What stays human | Typical AI depth | Typical fit signal |
|---|---|---|---|---|
| Campaign trafficking and setup QA | Insertion orders are turned into campaign setups and checked against targeting, budget and creative specs before launch. | Unusual deals and client-facing changes. | Level 3 | Many similar campaigns, costly setup errors |
| Floor-price and yield adjustments | Floors are proposed from bid and win-rate data, applied within limits you set, and logged. | Yield strategy and limits. | Level 3 | Large volume, measurable outcome |
| Invalid-traffic (IVT) review | Rule-based scoring plus anomaly detection returns Clean / Suspicious / Blocked on the bid path. Our work: see the IVT detection case study. | Threshold setting and review of disputed verdicts. | Level 3 | Real-time decisions, clear signals |
| Campaign reporting and performance questions | Plain-English questions are answered from live metrics and campaign documents, with every answer traced to source. Our work: see the analytics copilot case study. | Interpretation for clients and commercial decisions. | Level 2 | Repeated ad hoc report requests |
| Bid, clear and bill reconciliation | Every bid state change passes a check before it commits, so discrepancies are prevented, not found later in reconciliation. | Commercial disputes and rule changes. | Level 4 | Recurring discrepancies, slow reconciliation |
| Media discovery and negotiation | Buyer and seller agents discover audiences, negotiate terms and activate campaigns, with every step logged. Our work, at proof-of-concept stage: see the AdCP case study. | Deal approval, pricing strategy and relationships. | Level 4 | Many small deals, slow manual negotiation |
Evidence gathered as work happens, not reconstructed later.
Reporting and compliance work is often a person copying data between systems so someone else can check it. Much of that can be produced and logged automatically. See also finance operations and professional services.
| Workflow | What AI-native looks like | What stays human | Typical AI depth | Typical fit signal |
|---|---|---|---|---|
| Recurring management reporting | Weekly and monthly packs are assembled from source systems, with changes explained and numbers traceable. | Commentary, priorities and decisions. | Level 2 | Same report, rebuilt by hand each period |
| Data quality checks on incoming feeds | Incoming files and feeds are validated against schema and business rules before they reach downstream systems. | Rule definitions and fixing root causes with data owners. | Level 1 | Frequent feeds, known failure patterns |
| Compliance evidence collection | Evidence for controls is gathered from logs and systems on a schedule and filed against each requirement. | Control design and attestation. | Level 2 | Recurring audits, scattered evidence |
| Contract review against a playbook | Incoming contracts are compared with your standard positions. Deviations are highlighted with suggested fallback wording. | Negotiation, risk acceptance and legal sign-off. | Level 2 | Standard templates, many similar contracts |
| User access reviews | Access lists are pulled from each system, compared with roles and flagged where they do not match. | Approval and revocation decisions. | Level 2 | Many systems, periodic review cycle |
The same workflows, seen through your sector.
Industry pages show which of these workflows usually matter most in each sector and what we would examine first.
| AdTech & programmatic | Trafficking, yield, IVT, reconciliation and reporting. The sector where our own production work sits. |
| E-commerce & marketplaces | Order exceptions, returns, catalogue data and supplier follow-up. |
| SaaS & technology | Lead routing, onboarding, support triage and renewal monitoring. |
| Finance operations | Invoice matching, collections, reconciliation and compliance evidence. |
| Professional services | Proposals, contract review, reporting and client onboarding. |
Pick one workflow → run the free diagnostic on it.
Choose the workflow from this page that costs your team the most time. The free diagnostic takes 10 short steps, about 8–12 minutes, and returns an Operational Leverage Snapshot: where the leverage signals are, how deep the AI change could go, and what still needs measuring. It does not promise savings. If the evidence is strong, the next step is a fixed-fee Operational Leverage Audit.