Wiki:Presentations/Adaptive
From WFM Labs
| Presentation source | |
|---|---|
| ID | PS-001
|
| Origin | SWPP 2026, 44 slides |
| Blocks | 22 |
| Extracted | 20 August 2026 |
| Method | CP-CNT-001 |
The argument from Adaptive: Building Workforce Systems for an Unpredictable Future, held as markdown blocks so decks can be built from it against any template.
No artwork, no layout, no branding — see Wiki:Presentations for the index and the method.
Before building from this
Three things in the source material need attention before they meet an audience that checks numbers.
- The four correction factors in C21 carry no source. Rebound demand, escalation tax, complexity concentration and savings erosion are the most decision-relevant content here and the least affordable to have challenged. Source them, or present them as calibrated estimates with the reasoning shown.
- The 15/25/60 value split in C18 is likewise unsourced. Treat as illustrative unless evidenced.
- C13 names vendors in each ecosystem quadrant. Appropriate at a practitioner conference; strip them for an internal corporate audience, where it reads as procurement advice.
Two smaller ones: C7 calls level two "precision theatre", which is criticism of the workforce management function and lands differently outside a practitioner room. C14 inherits a terminology inconsistency — the collaborative staffing model appears as N-star cognitive portfolio, N-star cognitive load and N* cognitive portfolio model. Settle one.
The only externally sourced figures in the original are the Sequoia quotation in C5 and the Gartner rehiring figure in C4.
The elephant
The commercial reality, named before anyone else names it — and the reframe that lets a room hear it.
## C1 · The pace of change is itself the change
**CLAIM** Each era of work has arrived faster than the last, and the current one arrived in under
five years.
**WHY** Sets up everything that follows. If eras used to last millennia and now last years, no plan
built for a steady state survives.
**EVIDENCE** Stone Age (10,000+ BC, millennia) → Bronze/Iron (3500 BC–500 AD, ~4,000 years) →
Industrial (1750–1950, 200 years) → Information (1950–2020, 70 years) → **Collaborative
Intelligence (2020–present, under 5 years)**. Axis: pace of change.
**FORM** Sequence, compressing. Five or six stages, where the compression is the visual point.
**NOTES** Works as an opener because nobody argues with it. Do not spend more than one slide.
---
## C2 · Service has re-platformed five times, and the sixth is agent-to-agent
**CLAIM** Every service era redefined the channel. This one redefines the worker.
**WHY** The distinction between "another channel" and "a different kind of worker" is what most
planning gets wrong.
**EVIDENCE** Branch 1950–75 ("walk in") → Phone 1975–95 ("1-800 number") → Internet 1995–2010
("self-service + email") → Omnichannel 2010–25 ("always connected") → AI Wave 2025–now ("AI does the
work") → **Personal Agent, emerging ("agent talks to agent")**.
**FORM** Sequence, six stages with dates.
**NOTES** The verbatim line — *"each era redefined the channel; this one redefines the worker"* — is
the payload. The timeline is the setup.
---
## C3 · The elephant in the room
**CLAIM** If a company could run its contact centre with 1% of the workforce and hold the same value
delivered, it would — tomorrow.
**WHY** Names the commercial reality before anyone else does, which is what earns the right to
complicate it afterwards. Refusing to say it is what makes workforce leaders sound defensive.
**EVIDENCE** A descending staircase: **10,000 → 5,000 → 2,500 → 1,000 → 500 → 100**, under an arrow
labelled PRESSURE.
**FORM** A descending progression. Six steps. One headline statement carries it.
**NOTES** Deliver it flatly. It is not cynicism and should not be performed as such — it is the
premise the rest of the argument answers.
---
## C4 · The headlines are real, and so is the correction
**CLAIM** The market rewards headcount reduction immediately and re-hiring follows quietly.
**WHY** Both halves matter. The pressure is real; so is the overcorrection.
**EVIDENCE** A named company went 10K to 6K with stock +20%; the S&P software index −20%; only 20%
of firms reduced staffing; **50% will rehire by 2027 (Gartner)**.
**FORM** Three or four parallel data points.
**NOTES** The Gartner figure is the only externally sourced number in the source deck. Use it, and
be careful about presenting the others as equivalent.
---
## C5 · Talkers became doers in 2025
**CLAIM** The 2023–24 generation of AI talked; the 2026–27 generation does the work.
**WHY** Explains why prior automation waves under-delivered and this one behaves differently — which
pre-empts the "we've heard this before" objection.
**EVIDENCE** 2023–24 *"applications that talk. Chatbots. Impact was limited."* · 2025 *"the
transition year. From talking to doing. The repricing begins."* · 2026–27 *"applications that do.
Agents, not chatbots."* Quote: *"The AI applications of 2023 and 2024 were talkers. The AI
applications of 2026 and 2027 will be doers."* — Sequoia Capital.
**FORM** Three eras, one quotation carrying the slide.
**NOTES** Sourced. Safe to lean on.
---
## C6 · The elephant has two sides
**CLAIM** Script-following work shifts to AI; what remains is work worth doing.
**WHY** The honest reframe. Not "no jobs lost" — **fewer jobs, better jobs.**
**EVIDENCE** *"Fewer jobs. Better jobs."*
**FORM** Two-panel comparison, or a single statement.
**NOTES** This is the sentence that lets a room hear C3 without shutting down. Keep the two together.
Where you stand
A five-level maturity model, two lenses of readiness, and the diagnosis that matters: the gap.
## C7 · Five levels of workforce management maturity
**CLAIM** A five-level model calibrates where a planning function actually sits, and most sit at
level two.
**WHY** Establishes shared vocabulary before any diagnosis. Nobody can be told they have a gap until
there is an agreed scale.
**EVIDENCE**
- **L1 Initial / Manual** — supervisors monitor queues personally; spreadsheets; all-hands during
peaks. *Growth trigger: volume exceeds 100 agents.*
- **L2 Foundational / Traditional** — professional platform, structured cycles, skills-based
scheduling. *"Forecast → Erlang → single number → defend it."* **"This is where precision theater
lives."** **85% live here.**
- **L3 Progressive / Variance Harvesting** — real-time execution, continuous intraday optimisation,
*variance equals normal, not failure*.
- **L4 Advanced / Where the Value Model Lives** — evergreen planning, multi-objective across cost,
CX and EX, Monte Carlo and probabilistic, enterprise integration. **The phase transition sits at
L3→L4: different math, different skills, different equation.**
- **L5 Pioneering / Enterprise-Wide Intelligence** — distributed intelligence, human-AI
orchestration, autonomous operations. *No one is fully here yet.*
**FORM** Five-stage progression, with two annotations — where most sit, and where the phase
transition is.
**NOTES** Needs at least two slides: the model, then the levels. **"Precision theatre" reads as
criticism of the function** — fine at a practitioner conference, risky in a corporate room.
---
## C8 · Two lenses of readiness, and most assess one
**CLAIM** Organisational readiness and AI deployment are separate diagnoses and both are needed.
**WHY** A team can be process-ready and AI-blind, or the reverse. One lens misses it.
**EVIDENCE** Lens 1 — Process Complexity × Technology Architecture, giving *Fast Track · Faster Than
Expected · Infrastructure Gap · Longest Road*. Lens 2 — Customer-Facing AI × Agent-Facing AI, giving
*Exploring · Customer-First · Agent-First · Integrated*. Footer: *"Most organizations assess one. You
need both."*
**FORM** Two 2×2s, or one slide per lens.
**NOTES** Only use a 2×2 where both axes are genuinely measured. These are.
---
## C9 · The gap is the risk
**CLAIM** The risk is not the maturity level. It is the distance between company speed and planning
capability, and whether that distance is widening.
**WHY** Converts a self-assessment into an argument for action. A level-two function in a slow
company is fine; the same function in a fast one is a liability.
**EVIDENCE** Three assessments converge: WFM maturity → organisational readiness → AI deployment,
into **"THE GAP = company speed minus WFM capability."** Then: *"If the gap is growing, your
planning function is falling behind."* And: **"The company will not slow down for WFM to catch up."**
**FORM** Three inputs converging to one, then a two-panel contrast.
**NOTES** The strongest structural idea in the whole deck. Everything before it is setup.
The fork
The strategic choice, and what has to be rebuilt.
## C10 · The fork is strategic, not technical
**CLAIM** Keep optimising the old equation and be commoditised, or move to a value framework and
hold the seat at the table.
**WHY** Forces a choice rather than offering a menu. And names it correctly — this is not a tooling
decision.
**EVIDENCE** **Old equation:** volume · handle time · shrinkage · occupancy · attrition →
*"increasingly automated, increasingly commoditized."* **Value framework:** value tiers · routing
models · human-AI optimisation · probabilistic planning → *"strategic, indispensable, the seat at the
table."* Footer: **"This is not a technology decision. It is a strategic one."**
**FORM** Two paths, opposed, with the choice named.
**NOTES** The hinge of the deck. Everything after it is the answer.
---
## C11 · Three things change
**CLAIM** Systems, the workforce model, and the planning model all have to be rebuilt — not one of
them.
**WHY** Prevents the audience taking the easy third and leaving.
**EVIDENCE** *"Your systems — the adaptive ecosystem / Your workforce model — three pools, three
staffing models / Your planning model — from volume to value."*
**FORM** Three parallel items. Reused as a section divider, highlighting each in turn.
**NOTES** Structural. Cheap and it holds the back half of the deck together.
Systems and the workforce model
Two of the three things that change.
## C12 · Monolith versus best-in-breed
**CLAIM** One vendor means one roadmap and one ceiling. The alternative is specialised components
exchanging data through open interfaces.
**WHY** Most planning functions are constrained by their platform's release cycle and do not name
it as a constraint.
**EVIDENCE** Monolith — capacity planning, WFM core, automation and analytics locked in one cube:
*"one vendor, one roadmap, one ceiling."* Best-in-breed — four components around an API hub:
*"specialized, interchangeable, open."* Footer: **"This does not mean rip and replace. It means
augmenting what you already have."**
**FORM** Before/after architecture comparison.
**NOTES** That footer is what stops it being heard as a replacement pitch. Do not drop it.
---
## C13 · One adaptive ecosystem, four capabilities
**CLAIM** The future is not one tool. It is four capabilities connected by architecture that lets
data move.
**WHY** Names what actually has to be built, and the connective tissue is the part that gets
underfunded.
**EVIDENCE** Four quadrants around an API hub: **capacity planning · WFM core · real-time automation
· analytics.** Footer: *"Strong architecture with open data exchange amongst layers. That is the
foundation."*
**FORM** Four components around a centre, with bidirectional flow.
**NOTES** The source deck names vendors in each quadrant. **Strip vendor names for any internal
corporate audience** — it reads as procurement advice and invites a different argument.
---
## C14 · Three pools, three staffing models
**CLAIM** Work divides into three handler classes, and **one formula cannot serve all three.**
**WHY** This is the workforce-model half of the answer, and the mathematics genuinely differs — it
is not a labelling exercise.
**EVIDENCE**
- **AA — agentic AI.** Staffed on a *transaction cost model*.
- **Collab — human plus AI oversight.** Staffed on an *N-star cognitive portfolio model*.
- **Specialist — expert human.** Staffed on *Erlang plus a complexity premium*.
Admission threshold to Specialist: *AI capability under 30 percent, or value score 8 plus.*
**FORM** Three parallel items, each with a distinct attribute.
**NOTES** **Settle the terminology before use** — the source deck calls the middle model *N-star
cognitive portfolio*, *N-star cognitive load* and *N\* cognitive portfolio model* on three different
slides.
---
## C15 · The oversight role does not exist yet
**CLAIM** A human overseeing five to seven concurrent AI sessions is a fundamentally new role, and
almost nobody staffs for it.
**WHY** It is the pool that gets invented project by project and designed nowhere. It also changes
what gets recruited and where those people can sit.
**EVIDENCE** *"Humans oversee 5-7 concurrent AI sessions."* · *"Intervene on judgment calls."* ·
*"Approve escalation decisions."* · **"This role does not exist in most organizations today."** ·
**"Not queue-based. This is oversight. A fundamentally new role."** Framing: **air traffic control,
not call centre.**
**FORM** One role, its responsibilities, and the ratio stated large.
**NOTES** *"Air traffic control, not call centre"* is the line people remember. Also: **"AI-assisted,
not AI-free. Agents have AI tools. But the human IS the product."**
The planning model
The third, and the destination. The Value Planning Model and its components.
## C16 · The trap: assuming deflected volume is gone
**CLAIM** The old carve-out — total volume minus what automation absorbs, plan for the remainder —
breaks when the automation boundary is ragged and moving.
**WHY** This is the specific error the value model corrects, and stating it first is what makes the
model land as a fix rather than as a framework.
**EVIDENCE** Old carve-out: circle minus wedge, *"it worked well enough."* Agentic reality: a ragged
boundary leaking four ways — **rebound demand · escalation tax · complexity concentration · new
capabilities.** *"The pool is never static."* **THE TRAP: assuming automated volume is gone and you
never need to plan for it again.** → *"Start with the value of every interaction, not the volume you
hope to deflect."*
**FORM** Before/after, where the "after" is deliberately untidy.
**NOTES** The ragged boundary is the visual argument. A clean diagram here defeats the point.
---
## C17 · The value of an hour
**CLAIM** Not all working time is equal, and four tiers separate what is worth protecting from what
is worth removing.
**WHY** The foundation of value-based planning. Without it, every hour is treated as interchangeable.
**EVIDENCE**
- **T1 Doing the work** — customer-facing productive. Calls, chats, claims, back-office fulfilment.
**High value.**
- **T2 Getting ready** — training, coaching, development, team meetings. **Positive.**
- **T3 The cost of being human** — breaks, lunches, PTO. **Neutral.**
- **T4 The leak** — unaccountable time, idle without assignment. **Negative.**
An **automation zone** straddles the T1/T2 boundary.
**FORM** Four tiers, ranked, with one boundary marked.
**NOTES** T4 named as *"the leak"* does a lot of work. Keep the word.
---
## C18 · Not all Tier 1 is equal — and the split
**CLAIM** Inside productive customer-facing time there are three value bands, and one formula treats
them identically.
**WHY** The blind spot that makes headcount planning wrong. Balance checks and retention calls are
not the same work.
**EVIDENCE** **High** — retention, complaints, expert judgment. **Medium** — disputes, billing,
AI-assisted. **Low** — balance checks, status, *"pennies per transaction."* *"One formula treats them
all the same."*
Worked split: **High 15% → human only · Medium 25% → collaboration · Low 60% → AI only.** Footer:
**"You have not reduced headcount. You have increased the value of every human hour."**
**FORM** Magnifier into three bands, then a three-way split with percentages.
**NOTES** The 15/25/60 carries no source in the original. **Relabel as an illustrative split, or
source it, before it goes in front of an audience that checks.** That footer is the most reusable
sentence in the deck.
---
## C19 · The Value Planning Model
**CLAIM** Replace *volume minus deflection, planned with Erlang* with *total workload, classified by
value, routed by design, staffed by handler class, measured on value per dollar.*
**WHY** The named model, and the deck's destination. Its structural move is that **deflection is
deleted as a step** — total workload is never reduced up front, and "channel" is redefined as
handler class rather than media.
**EVIDENCE**
**The old way:** inbound volume (*all inquiries by channel*) → deduct deflection (*IVR, bots,
self-service*) → plan with Erlang (*offered volume to agents, single formula*) → **headcount.**
**The new way:** total workload (*all inquiry types, all channels*) → **value classification**
(*high / medium / low*) → **routing decision** (*human / collaborative / AI*) → **staff by channel**
(*human, collaborative, AI — not chat, email, phone*) → **value output** (*spend vs value derived
per dollar*).
Footer: *"Same workload. Right handler for the right work. Higher value per dollar spent."*
**FORM** Two contrasted sequences, old above and new below. Three stages against five.
**NOTES** **The single most important slide.** Worth real design effort. The three new stages should
be visibly marked as new.
---
## C20 · Value first, then route
**CLAIM** Routing is two-dimensional — **value × AI capability** — not a single score.
**WHY** A one-dimensional score sends low-value work to AI that cannot do it, and holds high-value
work AI could assist with. Both axes are needed.
**EVIDENCE** Three stages. **Classify by value** — interaction types scored 1–10. **Route by design**
— *high value or low AI capability* → Specialist; *medium complexity, AI-assistable* → Collab; *low
value, high AI capability* → AA. **Staff by model** — Erlang plus complexity premium · N-star
cognitive portfolio · transaction cost.
A low-scoring item still routes to Specialist if AI capability is low. The same score can land in two
different bundles.
**FORM** Three-stage mechanic, with the two-axis rule made explicit.
**NOTES** The two-dimensionality is the whole point and is easy to lose in the visual.
---
## C21 · Why AI planning is different
**CLAIM** Containment does not convert to savings one-for-one. Expect roughly half.
**WHY** The direct answer to any headline automation target, and the most decision-relevant content
in the deck.
**EVIDENCE** Four second-order effects — **rebound demand** (*volume grows 20%*) · **escalation tax**
(*AI failures reach humans at 5–10× original cost*) · **complexity concentration** (*each 10%
automation raises remaining AHT by 6.5%*) · **savings erosion** (*year 1: 50–70% realised; year 3:
only 25–45% remains*). Footer: **"50% containment does not mean 50% savings. Expect 25-35%."**
**FORM** Four items, one of which is the conclusion.
**NOTES** ⚠ **None of these four figures carries a source in the original.** They are the most useful
material in the deck and the least affordable to have challenged. **Source them, or relabel as
calibrated estimates with the reasoning shown, before any audience that checks.** Presenting an
estimate as a finding is the one avoidable failure.
The close
## C22 · The strategic choice
**CLAIM** The question is not whether workforce management survives. It is whether you lead the
transformation.
**WHY** Closes on agency rather than threat, which is what makes people act.
**EVIDENCE** *"Where do you want to sit in five years?"* **Stay in the queue** vs **own the
ecosystem.** Closing thesis: **"The future of workforce management is not about counting heads. It is
about maximizing value."**
**FORM** Two paths, then one statement.
**NOTES** End here. Do not add a summary slide after it.
