Allocation by availability is not allocation by capacity

"We need Sam on this—Sam is free Thursday." Thursday availability ignores that Sam already carries forty hours of open work across two releases, three approval queues, and the onboarding runbook only Sam knows.

AI optimal resource allocation does not mean an algorithm perfectly solves staffing NP-hard problems. It means visibility before assign—heatmaps, portfolio views, estimates compared to logged time—and AI that narrates overload patterns so managers redistribute with evidence.

WKFGo's honest MCP tools for allocation:

workload_heatmap — Visual concentration of open work per person
portfolio_overview — Cross-project conflicts when the same expert is double-booked
estimate_task + time logs — Planned vs actual load on upcoming work

A four-step allocation framework

1. Baseline capacity honestly

Know productive hours per week—not calendar hours. Subtract meetings, on-call, PTO. AI cannot fix fantasy capacity inputs.

2. Heatmap before every assign

Before assign_task, check workload_heatmap. If the row is already red, the conversation is redistribute—not "can you squeeze this in?"

3. Portfolio conflict check

Experts span projects. portfolio_overview reveals when one person is critical path on three milestones the same month.

4. Pair with estimation and flow

estimate_task on incoming work—compare to remaining capacity
flow_aging on their queue—overload often shows as stuck WIP, not missing hours

How AI assists—not replaces—staffing judgment

Use MCP prompts:

"Show workload_heatmap—who is red this sprint?"
"List tasks assigned to Alex across all projects with estimates."
"Given heatmap, who has capacity under twenty hours open?"
"Summarize portfolio_overview conflicts for mobile release."

AI returns tool output; managers negotiate with leads and assign_task after agreement.

Role-specific briefs help leadership staffing forks:

executive_brief role=chro — Capacity and burnout signals
simulate_scenario mode add_people — Model adding contractors vs delay

Report data gaps when time logs are missing—do not invent utilization percentages.

Anti-patterns

Spreadsheet capacity updated monthly while board changes daily
Assigning to the "fastest" person every sprint until they quit
Ignoring approval load as non-work
AI staffing plans with no names tied to real assignee IDs

Manager playbook: weekly five minutes

Open workload_heatmap in sprint planning
Flag any row red two weeks running
Redistribute or defer before adding new scope
Document backup owners for key-person rows (search_people + wiki)

Sustained red heatmap + missed PTO + overtime logs = conversation before exit interview.

Connecting allocation to finance and schedule

Overstaffing and understaffing both cost money:

finance_summary when contractor add is the fix
Release compression requires heatmap check—not only Gantt date moves
decision_inbox when staffing fork needs executive budget approval

Skill matrix vs heatmap quantity

workload_heatmap answers "how much," not "how hard." Two tasks with equal estimates may differ wildly in risk—production migration vs copy tweak. Leads should maintain lightweight wiki runbooks (update_wiki_page) marking critical skills so AI search_people and human staffing dialogs complement heatmap quantity.

Contractor and vendor allocation

When simulate_scenario mode add_people is on the table, pair heatmap evidence with finance_summary contractor burn. Adding bodies without budget fork in decision_inbox repeats the "heroics then invoice surprise" pattern.

Onboarding and bus factor

If heatmap shows one row green only because one person holds tacit knowledge, allocation risk is key-person dependency—not healthy spare capacity. Document backup assignees on wiki; assign_task to trainees for low-risk work before crisis.

Read path for new PMs

New PMs should run workload_heatmap before their first sprint planning meeting—not after overload complaints arrive. Document the ritual in project wiki so AI assistants can remind leads via MCP when heatmap was last checked.

Resource allocation without portfolio context assigns heroes to every fire. Before assigning a critical path task, check portfolio_overview for cross-project load on the same expert. AI can narrate trade-offs—who gains capacity if we defer Project B—but humans assign_task after reviewing heatmap evidence, not after accepting the first suggestion.

Run a monthly audit: do AI answers cite task IDs from MCP? If not, tighten prompts to require citations before any stakeholder-facing draft leaves the team.

Permission-aware tools mean the same question gets different grounded answers for PM vs contractor—that is a feature, not a bug. Train teams to expect scoped truth.

FAQ — AI resource allocation

Does workload_heatmap replace HR systems?
No—it shows delivery load on tasks; HR owns headcount and hiring pipeline.

Can AI auto-assign tasks?
assign_task exists; use after human review—especially on committed sprint items.

Remote/async teams?
Heatmap uses open assignments and estimates—ensure time logging culture for accuracy.

What about skill fit?
Heatmap shows quantity; leads still judge skill—use wiki and search_people for expertise signals.

Try it now

Put these patterns on live project data—not slide decks.