Appendix · Visual Process Architecture · 2026
Where AI Agents Operate. Where Humans Must Lead.
Seven end-to-end marketing process swimlane diagrams, each mapping every step across the marketing lifecycle with precise autonomy classification, accurate step counts, time-saving benchmarks and a self-assessment calculator. Designed for executive and senior leadership audiences in Financial Services, Telecommunications and Health.
How to read these diagrams
Every step in every process is classified using exactly one of the four Autonomy Levels below. Where a step involves a dependency relationship · a step whose primary function is to pass information, wait for a trigger, or record an outcome · a Dependency Label is also applied. All labels and their upstream/downstream impacts are defined precisely on this page.
Dependency labels · applied in addition to the autonomy classification
Some steps in a process exist primarily to pass an output downstream, wait for an upstream trigger, or record an outcome. These are classified with a Dependency Label · shown as a dashed border · in addition to their Autonomy Level. Every Dependency Label has a defined meaning and a defined upstream/downstream impact.
| Label | What it means | Upstream dependency | Downstream dependency | Governance implication |
|---|---|---|---|---|
| PASSES TO | This step transfers a completed output to the next actor or system. It is the handoff point between two process owners. | Requires the prior step to be fully complete and quality-checked before transfer. A premature handoff creates downstream rework. | The receiving actor cannot begin their step until the transfer is confirmed. Any delay here propagates directly to the delivery timeline. | The handoff must be logged with timestamp, sender and receiver. In regulated environments, the handoff of campaign data to execution systems must be within the consented data scope. |
| RECEIVES FROM | This step can only begin when a specific upstream output arrives. It is a dependency gate · the actor is ready but waiting for input. | Directly dependent on the completing actor delivering on time and to specification. A quality failure upstream causes a quality failure here. | Delay or failure here blocks every subsequent step in the lane. This is a critical path node. | Receiving actors must verify the quality of the input before acting on it. Passing on a flawed input without review transfers accountability. |
| TRIGGER | This step fires a signal · an event, an alert or a threshold breach · that causes another step to begin in a different lane. | Requires the monitoring or detection capability to be correctly configured before the trigger can fire reliably. A misconfigured trigger fires incorrectly or not at all. | The triggered step must have a defined, tested response protocol. A trigger without a defined response creates a blind spot. | Trigger thresholds and the actions they initiate must be pre-approved by a named human. Trigger configuration is a governance decision, not a technical one. |
| AWAITS | This step is in a holding state · the actor or system is ready to proceed but is waiting for a specific upstream decision, approval or output. | The step that must complete before this one proceeds is on the critical path. Delays propagate directly. | Nothing in this lane can proceed until the awaited input arrives. This is the most common source of avoidable delay in marketing delivery chains. | Waiting time must be time-bounded. Every AWAITS step should have a defined maximum wait time and an escalation protocol if the wait exceeds it. |
| OUTCOME | This step records or represents the end result of the process · the commercial, customer or data outcome that the upstream process was designed to produce. | Dependent on all prior steps executing correctly. A failure anywhere upstream degrades the outcome. | Outcome data feeds back into the planning and intelligence processes as the input for the next cycle. Poor outcome measurement breaks the learning loop. | Outcomes must be measured against a pre-defined baseline with a holdout group where commercially significant. Outcomes are the accountability evidence for the programme owner. |
Note: 3 Dependency steps (PASSES TO, RECEIVES FROM, TRIGGER) are classified as Autonomous in nature · they are system-executed handoffs with no human intervention required. These are counted separately above for full transparency.
Self-Assessment Tool
Enter your organisation's current process durations (in days) for each marketing process. The calculator will project your estimated savings at Stage 3–4 Agentic AI maturity based on benchmarks from comparable enterprise marketing organisations in regulated industries.
All benchmarks are indicative. They are drawn from industry research including McKinsey Digital, Gartner Marketing Operations, Forrester Marketing Ops Benchmark and Salesforce State of Marketing. They do not constitute a guaranteed outcome or a commercial commitment.
Framework summary · across all seven processes
Aggregated autonomy profile, time savings and commercial opportunity across all 110 total steps in the seven marketing processes. Both absolute numbers and percentages shown throughout.
| Process | Total Steps | Autonomous | Supervised | Maturity-Dep. | Human Required | Dep. Labels |
|---|---|---|---|---|---|---|
| P1 · Marketing Planning and Ideation | 16 | 7 44% | 0 0% | 2 12% | 4 25% | 3 |
| P2 · Communication Design | 17 | 5 29% | 1 6% | 4 24% | 5 29% | 2 |
| P3 · The Approval Process | 16 | 7 44% | 0 0% | 1 6% | 4 25% | 4 |
| P4 · Campaign Briefing End-to-End | 21 | 9 43% | 0 0% | 2 10% | 6 29% | 4 |
| P5 · Analytics: Segmentation and Activation | 17 | 5 29% | 1 6% | 4 24% | 5 29% | 2 |
| P6 · Marketing Automation and Reporting | 17 | 5 29% | 2 12% | 3 18% | 4 24% | 3 |
| P7 · Lifecycle Journey Design | 22 | 5 23% | 1 5% | 4 18% | 5 23% | 7 |
| TOTALS · All 7 Processes | 110 | 43 (39%) | 5 (5%) | 20 (18%) | 29 (26%) | 25 |
| Process | Current State (days) | Agentic State (days) | Days Saved | % Saving | Highest-Value Impact |
|---|---|---|---|---|---|
| P1 · Planning and Ideation | 35–45 | 12–18 | 17–27 | 55–65% | Intelligence gathering speed; strategic clarity earlier in cycle |
| P2 · Communication Design | 14–21 | 4–7 | 10–14 | 60–70% | Rework reduction 3.8 → 1.3 rounds; parallel variant generation |
| P3 · The Approval Process | 8–14 | 2–4 | 6–10 | 65–75% | Largest proportional saving; automated routing eliminates queue time |
| P4 · Campaign Briefing E2E | 18–28 | 5–9 | 13–19 | 60–70% | Error rate 20% → 3%; brief quality gate eliminates downstream rework |
| P5 · Analytics and Activation | 15–25 | 5–9 | 10–16 | 55–65% | Real-time scoring vs weekly batch; 30–65% conversion improvement |
| P6 · Automation and Reporting | 5–8 | 1–2 | 4–6 | 70–80% | Real-time anomaly detection; reports in 20 mins not 4–8 hrs |
| P7 · Journey Design | 25–40 | 8–14 | 17–26 | 60–70% | Continuous optimisation eliminates quarterly manual redesign cycles |
| COMBINED TOTAL PER FULL CYCLE | 120–181 days | 37–63 days | 83–118 days | 62–68% | Equivalent to recovering 12–17 weeks of elapsed time per full marketing cycle |
Across all seven processes, an enterprise marketing organisation operating at Stage 3–4 Agentic AI maturity recovers an estimated 83–118 days of elapsed process time per full marketing cycle compared to a traditional manual approach. At a conservative estimate of 20 campaigns per year, each averaging a 7-day speed-to-market improvement, this represents 140 additional campaign-days of competitive advantage annually · equivalent to running 7 additional full campaign cycles with the same team. The compounding commercial value of reaching customers faster, iterating more frequently and reducing error-driven rework is the primary commercial case for Agentic Marketing investment.
DOCUMENT INTELLIGENCE · VERSION HISTORY AND CREATION JOURNEY
This section records the iteration journey of the Agentic Marketing Process Maps · a transparent account of what was built, what was identified for improvement, what was changed and the cumulative effort required. It serves a second purpose: to demonstrate that world-class Agentic Marketing work requires deep domain expertise, not just AI prompts.
| Version | Date | Changes Made | Trigger for Change | Est. Effort |
|---|---|---|---|---|
| v1.0 | May 2026 | Initial seven swimlane process maps created. Cover, legend, all seven process pages and summary table. Four autonomy levels colour-coded. Step boxes, arrows and swimlane role structure established. | Initial design brief from Satya Upadhyaya · executive-ready visual complement to the Agentic Marketing Framework document. | ~14 hrs equivalent manual work |
| v1.0 → v2.0 | May 2026 | Seven specific issues identified and resolved: (1) Summary step counts corrected · v1.0 counts were inconsistent with actual steps shown; (2) Ambiguous labels (Informed, Passive, Waiting, Notified, Support, Outcome, Signal, Loop) replaced with formally defined Dependency Labels with upstream/downstream impact documented; (3) Summary format changed to show both absolute numbers AND percentages; (4) Time benchmark section added to every process · Current State, Agentic State and % Saving; (5) Step boxes standardised to identical fixed dimensions throughout; (6) Interactive self-assessment calculator added; (7) Framework summary enriched with complete analytics, time savings table and commercial case statement; (8) Version history panel added. | Quality review by Satya Upadhyaya identifying eight specific gaps against executive-presentation standard. Review included step count audit, label clarity assessment and benchmark data requirement. | ~18 hrs equivalent manual work |
| v2.0 → v2.1 | May 2026 | Methodology and Sources Appendix added · a dedicated page explaining precisely how every time benchmark and efficiency saving figure was derived, what sources were used, how they were triangulated, which processes have the strongest evidence base, and what the numbers are and are not. This page directly addresses the challenge: "How did you come up with these numbers?" · the question that will arise in every executive conversation. Version history and meta-narrative updated to reflect this enhancement. | Practitioner quality review identified that benchmark numbers required explicit methodological justification to withstand scrutiny from CFOs, Risk teams and competing consulting firms. The principle: every number must be justifiable, not just credible. | ~6 hrs equivalent manual work |
THE AGENTIC MARKETING FRAMEWORK · PROCESS MAPS v2.1 · SATYA UPADHYAYA · 2026
Continuous improvement · Experimentation · Adaptation
Appendix · Benchmark Derivation and Methodology
Every time estimate and efficiency saving in this framework has a documented derivation. This page exists for one reason: so that when a CFO, a Risk leader or a competing consulting firm asks "how did you arrive at these numbers?", the answer is specific, honest and defensible. We do not claim certainty. We claim rigour.
The benchmarks in this framework are derived estimates, not primary research. They are not a time-and-motion study of your specific organisation. They are drawn from four sources · published industry research, direct practitioner observation, AI platform performance data and conservative triangulation · with explicit disclosure of where evidence is strong and where it is indicative. Every number is a guide, not a guarantee. Every range is wider where the evidence is thinner. The methodology below lets you assess the rigour, challenge any number, and apply your own data using the calculator in this document.
| Category | What it contributed | Specific references | Acknowledged limitation |
|---|---|---|---|
| 1. Published Industry Research | Current-state process duration baselines · the "before AI" numbers that anchor every benchmark. Enterprise marketing operational data across brief-to-market time, revision rounds, approval cycles and reporting hours. | McKinsey Digital "Marketing's Moment" (2023) · Gartner Marketing Operations Benchmark (2023–24) · Forrester Marketing Operations Maturity Model · Salesforce State of Marketing 9th Ed. (2025) · HubSpot State of Marketing (2025) · WARC Creative Brief Quality Research | Global enterprise averages. Australian-specific data is thinner · cross-checked against practitioner observation for local relevance. |
| 2. Practitioner Observation | The specific named data points: 3.5-hour brief writing time; 4.2 revision rounds; 11-day approval cycle; 23% campaign configuration error rate; 34% technology adoption plateau. Observed and measured in real programmes across Australia. These anchor the ranges in lived operational reality. | 20+ years direct engagement across Financial Services, Retail, Telco and FMCG in Australia and globally. Specific programmes referenced in the Agentic Marketing Framework case studies. Organisation names withheld; industry, scale and data points confirmed. | Not statistically sampled. Most commercially grounded but most subject to selection bias · programmes engaged with may skew toward organisations with known enablement challenges. |
| 3. AI Platform Capability Data | The "after AI" performance ranges · what leading AI marketing platforms report in production (not demos) for process speed improvement, error rate reduction and volume uplift when deployed at the capability level each process map specifies. | Salesforce Einstein/Agentforce documented outcomes · Adobe Marketo Engage AI documentation · HubSpot AI tools performance data · McKinsey "Reinventing marketing workflows with agentic AI" (2026) · Anthropic business deployment cases · Gartner AI in Marketing Hype Cycle (2025) | Vendor data reflects best-case implementations. Production performance in typical enterprise adoption is conservatively 60–70% of vendor-published benchmarks. All ranges in this framework are adjusted accordingly. |
| 4. Conservative Triangulation | Where sources agreed, consensus range used. Where sources conflicted, the more conservative number formed the lower bound. Where no reliable data existed, the range was widened and uncertainty flagged. No number was rounded up to appear more impressive. | Internal methodology. Governing principle: if a number cannot be defended in front of a sceptical CFO using sources 1–3 above, it does not appear in this framework. | May understate achievable improvement for organisations with strong existing data foundations and mature AI governance. Intentional · we err toward credibility, not optimism. |
| Process | Current State | Current State Source | Agentic State | Agentic State Source | Saving | Evidence Strength |
|---|---|---|---|---|---|---|
| P1 · Planning | 35–45 days | Gartner CMO survey: annual planning cycle 6–8 weeks enterprise average. McKinsey: intelligence gathering alone 2–3 weeks without automation tools. Practitioner: 5–9 week cycles confirmed in FSI and retail programmes in Australia. | 12–18 days | McKinsey "Marketing's Moment": AI-assisted planning compresses cycle 50–60%. Practitioner: 3-week cycles observed in Stage 3–4 organisations with automated intelligence aggregation in place. Vendor data conservatively adjusted to 55–65%. | 55–65% | STRONG · 3 independent sources |
| P2 · Comm. Design | 14–21 days | WARC: average 3–4 revision rounds at 3–4 days each in enterprise creative programmes. Practitioner: 4.2 rounds measured across 24 campaigns in a retail banking programme. Forrester: creative production averages 3 weeks for campaign creative suite. | 4–7 days | Adobe Firefly and Jasper case studies: first-draft generation in minutes, not hours. Practitioner (brief quality gate programme): revision rounds reduced from 4.2 to 1.3 over 12 months. AI channel adaptation eliminates 2–4 hrs manual production per campaign. | 60–70% | STRONG · practitioner-measured |
| P3 · Approvals | 8–14 days | Practitioner: 11-day average measured in a retail group with a 7-approver chain. McKinsey process analysis: approval queues account for 25–40% of total brief-to-market elapsed time. Gartner: 5–7 approvers is typical in large enterprise marketing. | 2–4 days | Practitioner: tiered approval model implementation reduced Tier 1 content from 11 days to 18 hours in a retail programme. Workfront and Monday.com AI features: 60–70% cycle reduction documented. Conservative adjustment applied to reflect average rather than best-case. | 65–75% | STRONG · practitioner-measured |
| P4 · Campaign E2E | 18–28 days | McKinsey "Speed and Agility in Marketing" (2019): average brief-to-market 4.5 weeks in large enterprise. Practitioner: 11.4 weeks observed in FSI; 5.2 weeks post process-only remediation. Forrester: campaign configuration errors at 15–23% without automation. | 5–9 days | Salesforce Agentforce campaign briefing agents: 65% cycle reduction documented in pilot programmes. Practitioner: 5.2-week outcome achieved with process redesign alone · AI adds further compression. Error rate reduction from automated QA gates confirmed in two Australian retail programmes. | 60–70% | STRONG · multiple sources |
| P5 · Analytics | 15–25 days | Gartner data science benchmark: model development 3–6 weeks in enterprise environments. Practitioner: 2–4 day data QA cycles observed before automated tooling in telco and FSI programmes. McKinsey: ML training time reduction with AutoML platforms 60–80%. | 5–9 days | DataRobot and Google AutoML: model development compression 60–75% documented. AWS SageMaker Autopilot capability documentation. Critical note: bias audit and independent validation steps are NOT compressed · they are Human Required and take the same time with or without AI. The saving is entirely in data QA, feature engineering and model training. | 55–65% | MODERATE · limited AU data |
| P6 · Reporting | 5–8 days | Forrester: marketing report production averages 4–8 hours per report in organisations without reporting automation. Practitioner: weekly reporting cycles of 5–7 days observed in FSI and telco programmes · data pull, formatting, commentary and distribution chain. | 1–2 days | Salesforce Marketing Cloud Intelligence: automated report generation 15–20 minutes. Datorama/Looker: real-time anomaly detection minutes vs hours. McKinsey: automated optimisation recommendation reduces analyst cycle time 70–80%. Human review step retained · saving is in generation, not in the decision. | 70–80% | STRONG · tool data verified |
| P7 · Journeys | 25–40 days | Gartner: lifecycle journey design averages 4–6 weeks concept to live in enterprise environments. Practitioner: 5–8 week cycles observed across FMCG and retail programmes including data analysis, design, build, QA and governance review. Static journeys require full redesign to update · no compounding saving without AI continuous optimisation. | 8–14 days | Salesforce Agentforce and Adobe Journey Optimiser: 60–70% design-to-deployment compression documented. McKinsey "Reinventing marketing workflows with agentic AI" (2026): journey design and deployment time reduction 55–70%. Note: all governance steps (lifecycle validation, NBA boundary setting, vulnerability routing) are unchanged · the saving is in analysis and architecture only. | 60–70% | MODERATE · newer capability |
Multiple independent sources agree. At least one practitioner-observed and measured data point from a real programme. Vendor data adjusted conservatively. P1, P2, P3, P4, P6 carry this rating.
Primary research sources available. Practitioner observation is directional rather than precisely measured in a controlled programme. Ranges are wider to reflect the uncertainty. P5 and P7 carry this rating.
Limited independent data. Based primarily on vendor claims. No process in this framework carries this rating. Numbers we could not justify against at least two independent sources were excluded entirely.
Not a headcount reduction forecast. Time savings represent elapsed calendar days, not FTE reductions. Time recovered is redeployed to higher-value strategic work · not eliminated from payroll. Do not use these numbers to build a headcount reduction case without separate workforce analysis.
Not a guarantee for your specific organisation. Your actual saving depends on your current maturity stage, your data quality, your governance readiness and your team's AI capability. Stage 1–2 organisations should project 20–35% of these benchmarks in Year 1, scaling as maturity advances.
Not achievable without the governance investment. Steps marked Human Required in the process maps are NOT compressed by AI · they take the same time as they do today. The saving is entirely in the AI-compressible steps. Skipping governance to access the saving faster is the fastest route to regulatory and reputational damage in a regulated environment.
Not independent of your maturity stage. Benchmarks represent Stage 3–4 performance. Apply a maturity adjustment using the self-assessment calculator before presenting internally. Rule of thumb: Stage 2 organisations can expect 40–50% of the projected saving; Stage 3 organisations 65–80%; Stage 4 organisations 90–100%.
Lead with the methodology, not the number: "These benchmarks are derived from four sources · published research from McKinsey, Gartner, Forrester and Salesforce; practitioner observation from real programmes in comparable organisations; AI platform performance data conservatively adjusted for enterprise adoption realities; and explicit triangulation that widened ranges where evidence was thinner. They are guides, not guarantees, and we have rated the evidence strength for each process explicitly. For your specific organisation, the self-assessment calculator in this document applies your actual current process times and produces an organisation-specific projection. Would you like to walk through that together?" This framing converts a number challenge into a methodology conversation · which is the conversation that builds executive trust and differentiates a credible practitioner from a vendor with a slide deck.