Ad Hoc HR Projects (Task-Force Project Orchestration)¶
Questions or comments?
Post them as a comment on the tracking issue -- requires a GitHub account with access to this repo.
Novaworks AI | Super Agent orchestration (nova-superagent-agent) + x_novaw_platform
Major Change History¶
| Version | Date | Author / Notes |
|---|---|---|
| 1 | 2026-08-07 | Eswar Vandanapu — Feature Set created from a verbal brief on task-force-run ad hoc HR projects (e.g. upskilling initiatives). |
| 2 | 2026-08-10 | Eswar Vandanapu — Resolved the Codex-dependency question: template generation draws on the model's own knowledge plus whatever's already in the KB, not a dedicated curated corpus, which unblocks Feature 4 for this Feature Set's first Construction pass. Added Feature 5, an iterative agent that reviews and suggests changes to a drafted plan. |
| 3 | 2026-08-10 | Eswar Vandanapu — Resolved four more open questions: specs-advisor will own curating project-planning best-practice content as its own corpus (this Feature Set just depends on it); task forces can include anyone, though project initiation itself still needs a permission model; Projects aren't scoped to HR topics specifically, just weighted that way by HR Concierge execution; Feature 5's review is on-demand. Added design-partner evidence (Ateko) for the business case, and three new open questions raised by these answers. |
| 4 | 2026-08-10 | Eswar Vandanapu — Resolved who can initiate a project: it's a customer-configurable policy (HR only, HR plus Managers above a set level, or anyone), not a fixed rule — reusing the same feature-flag targeting pattern already built for guardrail policy configuration. Added System Admin to the persona table as the one who sets it. |
| 5 | 2026-08-10 | Eswar Vandanapu — Corrected the editor-generalization item: the Policy Editor's collaborative-authoring engine is confirmed scoped to KB/policy content only today, not an open question — it's now a stated dependency on specs-advisor (tracked on that spec's own Dependencies section), and no longer listed here as unresolved. |
| 6 | 2026-08-10 | Eswar Vandanapu — Resolved how Feature 5's suggestions reconcile with an already-agreed charter by making it its own Feature: Feature 6 reconciles and replans a project mid-execution, rather than leaving that behavior undefined inside Feature 5. |
| 7 | 2026-08-10 | Eswar Vandanapu — Added a Data Model (Conceptual) section (new in the concept.md template) covering Project, Task Force Member, Charter, and Task from Features 1–3. |
| 8 | 2026-08-10 | Eswar Vandanapu — Created the Commitment Meeting issue (#745) and Feature sub-issues (#746–748 for Features 1–3) in Novaworks-ai/Roadmap, milestone "Ideas." Ready for EL hand-off. |
| 9 | 2026-08-10 | Eswar Vandanapu — Flipped status from draft to ready — business case is solid, Features 1–3 are elaborated and ticketed. Candidate for the pilot's "ready concept.md" exit criterion, pending team review at the next standup. |
Business case¶
Novaworks' core differentiator over conventional HCM is that Nova agents execute work, not just recommend it, with human governance preserved throughout. This Feature Set extends that thesis to a kind of work Novaworks doesn't yet touch: a bounded, cross-functional initiative run by a group, not a single employee's transaction.
Today, when an HR leader or manager wants to run an ad hoc initiative — a company-wide upskilling program, a return-to-office rollout, a benefits open-enrollment push — there's no structured place to do it. The first two-plus weeks go to manually chasing task-force members over email and Slack just to agree on the initiative's goals and scope. Once the group finally settles on a plan, tracking who's doing what reverts to spreadsheets or a generic PM tool with no connection to HR's own data or agents.
The people who feel this are the Project Owner — an HR Ops practitioner, HRBP, or manager who decides to run the initiative — and the Task Force they assemble to help plan and execute it. The cost shows up as weeks of coordination overhead before any real work starts, initiatives that quietly stall once planning ends because no one's tracking execution, and every new initiative reinventing its own structure instead of building on what worked last time.
This isn't hypothetical. Design partner Ateko is already running one such initiative to handle an acquisition; a separate vision-demo project scoped for prospects is an upskilling program. Both are live examples of the ad hoc, task-force-run work this Feature Set targets — though Ateko's acquisition-handling initiative also raises a genuine question about whether an initiative at that scale needs more structure than this Feature Set's model provides (see Open Questions).
Customers get one place to describe a goal, assemble a task force, and carry an initiative from "here's what we're trying to do" through an agreed charter to a live, tracked set of distributed tasks — instead of stitching together an email thread, a doc, and a spreadsheet by hand. For Novaworks, this extends Super Agent's task-management capability from single-user transactions into multi-person, initiative-level orchestration — meaningfully larger territory for "the agent does the work" than anything shipped so far — and gives the collaborative-authoring investment built for the Policy Editor a second consumer outside Advisor, evidence that pattern generalizes as a platform capability rather than a one-off for policy content.
We intend to win this on two structural bets, not by matching a feature list. First, planning and execution live in the same agent stack that already runs HR's transactional work, so an agreed charter turns directly into distributed, tracked tasks rather than handing off to a disconnected PM tool. Second, charter agreement happens inside the same governed, real-time collaborative-authoring surface built for policy — proving that pattern generalizes beyond one content type. No HCM or adjacent HR platform combines task-force assembly, collaborative charter authoring, and agent-driven execution today (see Table-Stakes & Differentiation) — this is open ground, not a catch-up play.
Goal¶
Let a Project Owner turn a stated goal into a running, task-force-executed initiative — assemble the task force, collaboratively agree on a charter and goals, and have Super Agent generate and distribute the actual execution tasks — either from scratch, from a saved template, or generated from Novaworks' own accumulated knowledge and experience.
Scope¶
nova-superagent-agent — its existing task-management orchestration extends to cover both the planning and execution phases of a project. Likely new x_novaw_platform tables for Project, Task Force membership, and Charter (schema changes aren't drafted yet; that comes once this Feature Set moves further into Construction).
Projects aren't restricted to HR topics by design — a task force can route consult or execution tasks to Legal, Finance, or other functions. In practice they skew heavily HR because execution runs through Super Agent/HR Concierge, not because anything blocks a non-HR initiative.
This Feature Set depends on two capabilities it doesn't own, both from specs-advisor. The collaborative-authoring engine built for the Policy Editor is scoped to KB/policy content only today — generalizing it to a project-charter document type is a stated dependency on specs-advisor, tracked on that spec's own Dependencies section, not an open question on this Feature Set's side (see Related PRDs). Template generation (Feature 4) draws on the model's own general knowledge plus an opportunistic search of whatever's already in the KB today; a dedicated curated corpus of project-planning best practices is specs-advisor's Codex Curator role to build, as its own kind of content — this Feature Set consumes that corpus once it exists rather than curating it.
Who's allowed to initiate a project is a customer-configured policy, not a fixed rule — HR only, HR plus Managers at or above a set level, or anyone, depending on how a given customer wants to run this. Feature 1 builds this on the same feature-flag targeting pattern already in place for guardrail policy configuration (ai-guardrails-sensitive-routing/policy-configuration-console) rather than introducing a second mechanism for the same kind of decision.
Sources¶
- Lattice AI Agent — the closest competitive analog found; scoped to 1:1s, performance reviews, and individual goal-tracking, not cross-functional project coordination. Checked against this domain's competitor set (BambooHR, HiBob, Rippling, Workday, Darwinbox, Gusto) plus Lattice and 15Five, 2026-08-07.
Related PRDs¶
- specs-advisor: nova-policy-advisor / native-policy-editor-collaborative-authoring — the collaborative-editing engine this Feature Set's planning phase intends to reuse for charter co-authoring, built for KB/policy content and running via the Nova Advisor Agent's WebSocket layer. Confirmed not generalized beyond KB/policy content today; this Feature Set's Construction depends on
specs-advisorextending it, already logged as a dependency on that spec's own Dependencies section. - specs-advisor: nova-codex-curator — the curated-knowledge pipeline behind Advisor's policy content. Its corpus has nothing purpose-built for project-planning best practices today;
specs-advisorowns curating that as its own kind of content going forward, and Feature 4 will depend on it once it exists rather than this Feature Set curating anything itself. - coalition-of-the-willing — the same playbook this Feature Set follows: prove a pattern once, then treat it as reusable — there across integration types, here across domains.
- ai-guardrails-sensitive-routing: policy-configuration-console — the existing feature-flag targeting pattern Feature 1's project-initiation policy builds on, same repo, same mechanism, different policy.
Who this is for¶
| Persona / Role | Friction today | Value gained |
|---|---|---|
| Project Owner | No structured way to describe an ad hoc initiative's goal, pull together the right people, and get from "we should do this" to tracked execution — it's all manual coordination across email, Slack, and spreadsheets. | Describes the goal once, assembles a task force, and gets an agent-assisted path from charter through distributed, tracked tasks — optionally starting from a template or Novaworks' own accumulated experience. |
| Manager | Gets pulled onto ad hoc task forces with no shared planning surface beyond whatever the Project Owner improvises. | A governed, real-time collaborative space to help shape the charter and goals, then a clear, tracked set of tasks once planning concludes. |
| HR Operations | Often the Project Owner or a core task-force member for HR-led initiatives, with no platform-native way to run one distinct from day-to-day HR transactions. | The same task-management and collaborative-authoring investment the platform already has, extended to initiative-level work. |
| Employee | Can be pulled onto a task force as a subject-matter contributor (not just Manager/HR Operations), with no platform-native surface for that today. | The same collaborative planning and tracked-task experience as any other task-force member — initiating a project is a separate, customer-configured permission, not something every task-force member gets by default. |
| System Admin | No way to set who's allowed to initiate a project at all — it would otherwise default to "anyone," with no customer control. | Configures project-initiation eligibility (HR only, HR plus Managers at a set level, or anyone) using the same feature-flag targeting pattern already built for guardrail policy. |
| HRBP | Phase 2 — out of MVP scope. Upskilling and engagement initiatives are core HRBP territory, but this hasn't been scoped as an MVP persona yet. | — |
What we are building¶
| # | Feature | Folder |
|---|---|---|
| 1 | Project Creation & Task Force Assembly | project-creation-task-force-assembly/spec.md |
| 2 | Collaborative Charter & Goal Planning | collaborative-charter-goal-planning/spec.md |
| 3 | Agent-Assisted Task Generation & Distribution | agent-assisted-task-generation-distribution/spec.md |
| 4 | Project Templates & Knowledge-Assisted Generation | Spec not yet started |
| 5 | Agent-Assisted Plan Review & Refinement | Spec not yet started |
| 6 | Project Reconciliation & Replanning | Spec not yet started |
Data Model (Conceptual)¶
Covers only Features 1–3, the ones scoped so far. Features 4–6 (templates, on-demand review, reconciliation) will likely add more entities — a template/source reference, a review suggestion, a charter revision — once they're scoped; not drawn speculatively here.
erDiagram
PROJECT ||--o{ TASK_FORCE_MEMBER : "is assembled with"
PROJECT ||--|| CHARTER : "has"
PROJECT ||--o{ TASK : "generates"- Project — one ad hoc initiative, owned by a Project Owner.
- Task Force Member — a person (any existing employee, no role restriction) assembled onto a Project to help plan and execute it.
- Charter — the Project's agreed goal/plan; collaboratively authored, then finalized before tasks can generate from it.
- Task — a concrete piece of execution work generated from a finalized Charter and distributed to a Task Force Member.
Table-Stakes & Differentiation¶
| Table-stakes capability | Bar set by | Where we cover it |
|---|---|---|
| An AI agent grounded in company people data, usable in the flow of HR work | Lattice AI Agent | Feature 3 — though ours spans a multi-person initiative rather than 1:1/performance workflows |
| Multi-person project/task tracking | Generic tools (Asana, monday.com, Smartsheet) customers run alongside HR today | Features 1 & 3 — native to the same platform and agent stack as the rest of HR's data |
Differentiation. - Charter agreement happens inside a governed, real-time collaborative-authoring surface, not a doc or spreadsheet passed around by hand (Feature 2). - The agreed charter becomes actual distributed, tracked tasks through Super Agent's existing execution layer, closing the planning-to-execution gap generic PM tools leave the customer to bridge manually (Feature 3). - Optional template generation, drawing on the model's own knowledge and whatever's already in the KB rather than a blank page, gives customers a starting point competitors don't offer natively — with a dedicated curated corpus as a future upgrade path once one exists (Feature 4). - The plan doesn't stop improving once the charter is agreed: an agent reviews it on demand and suggests changes, something no generic PM tool or HR platform in this competitive set offers (Feature 5). - Execution doesn't have to stay locked to the original charter: when the plan needs to change mid-execution — whether from Feature 5's own suggestions or from real-world drift — Feature 6 reconciles the difference and replans, instead of leaving the task force to manually patch tasks around a stale plan. - No competitor combines all five — task force assembly, collaborative charter, agent-driven execution, ongoing agent-assisted plan review, and mid-execution reconciliation — in one product today.
UX & Interaction Design¶
Not yet available.
Open Questions¶
| # | Question | Blocks |
|---|---|---|
| 1 | Does an initiative at the scale of handling an acquisition need a heavier project type than this Feature Set's ad hoc Project/Charter/Task-Force model, or does the same shape flex to cover it? | The Goal/Scope boundary, and whether "ad hoc" undersells what this needs to support. |
For internal engineering review only. Not for external distribution.
Novaworks AI | Confidential