A Systems-Theoretic View of Hybrid Human-AI Teams
When I read management frameworks for engineering teams, I usually want to see the receipts. Not just "this works because experienced practitioners say so" — I want to see which choice maps onto which empirical or theoretical foundation. Most agile literature doesn't pass that test. The principles are sensible; the justification is mostly post-hoc.
The recent paper Management of organisations and teams with human and AI employees: A Systems-Theoretic Approach to the Honey Badger Framework (Fradelos, January 2026) takes the opposite approach. Each design choice is mapped onto a specific theoretical foundation with citations: agency theory, dynamic capabilities, stakeholder theory, behavioral economics, transaction cost economics, resource-based view. It's a useful lens even if you don't adopt the framework, because it forces the question: what's the actual mechanism by which a given practice is supposed to work?
This is worth working through, because hybrid human-AI teams are still rare enough that most operational decisions are being made on intuition. Intuition is fine; intuition with theoretical backing tends to fail less surprisingly.
The Six Theoretical Pillars
Strip the framework names and HBMF — and most defensible hybrid-team management approaches — sit on six theoretical foundations.
Agency theory
Agency theory describes the conflict between principals (owners/stakeholders) and agents (workers/managers) when their incentives diverge. In hybrid teams, this gets richer: there are multiple agents, and one of them is an AI assistant whose "incentives" are whatever the reward function or system prompt says they are.
The framework's response is centralized accountability through the Manager role, with the Guru as a structural counterweight that has C-level escalation rights. The mechanism is straightforward: explicit role separation reduces ambiguity about who owns what, which reduces the divergence in incentives.
The lesson for any hybrid team: the AI's "agency" is real even if it's not autonomous. If the AI is producing output that a human signs off on, the human's incentives — and the friction of signing off — affect what gets shipped. If you don't structurally separate the role that approves work from the role that owns delivery, you get rubber-stamping, which is the AI-era version of the agency problem.
Dynamic capabilities
Dynamic capabilities theory says competitive advantage comes from the ability to reconfigure resources rapidly in response to environmental change. In hybrid teams, this is what short, cancellable sprints are for: small batches preserve real-options value, and AI integration accelerates reconfiguration because the AI can take on new tasks faster than human reskilling.
The mechanism is: short cycle + AI flexibility = high reconfiguration velocity. The risk is the same as any high-velocity practice — you can reconfigure faster than you learn, which produces churn. The framework's answer is dashboard discipline: visible telemetry that catches reconfiguration that isn't producing learning.
Stakeholder theory
Stakeholder theory is why ESG isn't a separate compliance layer in serious frameworks for hybrid teams. The argument is: long-term success depends on aligning with all stakeholders, including the environment and the broader social context, and embedding that alignment into the operating model is more reliable than bolting it on at reporting time.
In hybrid teams specifically, the AI's energy footprint is a first-order ESG concern. So is the social effect of automating cognitive work that used to support junior career paths. Frameworks that don't think about this are not "ESG-neutral"; they're ESG-implicit, which usually means ESG-blind.
Behavioral economics
The mandatory weekly knowledge-gap declarations are the framework's behavioral-economics nudge. The mechanism is explicit: declaring what you don't know reduces the social cost of admitting it, which reduces knowledge hoarding, which improves cross-team learning rates.
This is one of the cleanest examples in the framework of a behavioral nudge with a mechanism behind it. Most "psychological safety" interventions in management literature are vague on mechanism. This one is specific: a weekly, public, low-stakes declaration of a gap reduces the marginal cost of admitting the gap during the rest of the week.
Transaction cost economics
Transaction cost economics is why the framework specifies its events in detail. Daily standups, sprint preparation, sprint review, stakeholder presentation — each is a structured information-flow event with defined inputs and outputs.
The mechanism: structured events reduce the transaction cost of information exchange, both within the team and at the team boundary. The risk is meeting bloat — more events with more structure can make information exchange more expensive, not less. The framework's answer is time-boxing and dashboard discipline: the events are bounded, and most information flow happens through the dashboard rather than through synchronous meetings.
For hybrid teams specifically, the AI assistant changes the transaction-cost calculus: routine information synthesis (sprint-end summaries, status updates, knowledge-gap analysis) can be handled by the AI at much lower cost than humans synthesizing the same information in meetings. This is a first-order operational improvement when done right.
Resource-based view
Resource-based view says competitive advantage comes from unique, inimitable, organization-specific resources. In hybrid teams, the inimitable resource isn't the AI assistant — that's commoditized — it's the integration of the AI into the team's specific workflow and the institutional knowledge of which problems the AI can and can't reliably handle.
This maps onto a practical observation: the value of AI in a team is heavily front-loaded into the integration phase. Two teams with the same AI tools and the same talent will produce dramatically different outcomes depending on how well the AI is integrated into their specific workflows.
What This Framework Reveals About Hybrid-Team Management
Stepping back from the specific framework, three things become clear when you look at hybrid-team management through a systems-theoretic lens.
The AI is a first-class systems variable
In most management frameworks, the AI is implicit — a productivity layer, not a system component. Once you treat it as a first-class variable, the systems behavior changes. The AI's reliability becomes a team performance metric. Its energy cost becomes a sustainability metric. Its failure modes become risk inputs. Its access boundaries become governance inputs.
This is the part most management literature still hasn't caught up with. Frameworks designed for all-human teams produce systematically wrong predictions about hybrid-team behavior because they treat AI as an environmental constant rather than as a system component.
Governance is the binding constraint
Across the six theoretical pillars, the consistent risk is governance failure. Agency divergence, reconfiguration without learning, stakeholder misalignment, hoarding under safety failure, transaction-cost inflation, integration without institutional capture — every one of them shows up when governance is weak.
This matches what I see in practice. Hybrid teams that are succeeding have invested heavily in process governance — not bureaucracy, but specific, low-overhead, high-leverage governance mechanisms (clear role boundaries, visible dashboards, mandatory audit cadences for AI-produced work). Teams that are struggling almost always have governance gaps in specific places.
Short cycles are an enabler, not a substitute
The seven-day cancellable sprint pattern is enabling rather than constituent. It enables rapid reconfiguration, rapid learning, rapid course-correction. It does not, by itself, produce them. A team running seven-day sprints with weak feedback loops, weak telemetry, and weak governance will run faster in the wrong direction.
This is why frameworks that focus only on cadence — "switch to two-week sprints," "switch to one-week sprints" — produce inconsistent results. The cadence is the enabler. The mechanism is the feedback and governance that the cadence makes affordable.
The Limits of the Framework
A systems-theoretic justification doesn't make a framework universally correct. Three honest limits:
Theoretical justification is not the same as empirical validation. Each pillar has citations to theory, but the integration of all pillars in a specific organizational context is not the same as the sum of the validated parts. (This is the "LEGO fallacy" the framework's own follow-up paper takes seriously.)
The framework assumes meaningful AI assistant capability. If your AI assistant produces unreliable output on the work it's assigned, the productivity and reconfiguration benefits don't materialize. Capability gaps in the AI become structural blockers in the framework.
Cultural and regulatory context matter. Behavioral nudges work in cultures where the nudge cost is acceptable. ESG-embedded practice works where ESG is a first-class organizational priority. Frameworks that work in Geneva or Barcelona may need adaptation in different cultural-regulatory contexts.
What to Take From This
Two practical takeaways for any CTO running hybrid teams:
-
Audit your management practices against the six pillars. For each operational practice, ask: what's the theoretical mechanism by which this is supposed to work? If you can't articulate one, the practice is intuition rather than design. Intuition might still be right, but it should be marked as such.
-
Pay disproportionate attention to governance. In hybrid teams, governance failures are the consistent failure mode across every theoretical pillar. The investment in low-overhead, high-leverage governance — clear roles, visible dashboards, mandatory cadences — pays for itself faster than any equivalent investment in tooling.
The systems-theoretic lens is useful even when you disagree with the specific framework. The exercise of asking "what's the mechanism" is the part that holds up.
Source: Fradelos, G. Management of organisations and teams with human and AI employees: A Systems-Theoretic Approach to the Honey Badger Framework (Geneva, January 5, 2026). SSRN 6306443.
If you're managing a hybrid human-AI engineering team and your management practices are running on intuition rather than design, talk to a CTO about deploying nearshore engineering capacity that's already operating in this regime.
