In this guide
Make room for both kinds of work
  1. 01Name the resource tension
  2. 02Choose an arrangement
  3. 03Connect decisions
  4. 04Plan the transfer
An organizing view for this guide. Exploration and exploitation may coexist; no universal structure or budget ratio is implied.

Organizational ambidexterity means being able to improve an established business while exploring opportunities that need different assumptions, routines and measures. The practical task is to give both kinds of work a viable home and make their dependencies explicit. Opening an innovation unit is one possible arrangement; it does not resolve the conflicts by itself.

This guide extends innovation strategy into an organizational question: who protects exploratory work, who keeps current commitments reliable, and how will a useful experiment become ordinary business?

Why the two kinds of work pull differently

James March’s exploration and exploitation paper describes a resource tension between searching for possibilities and refining what is already understood. Exploitation includes activities such as execution and efficiency; exploration includes experimentation and discovery. These are descriptions of work, rather than personality types.

O’Reilly and Tushman’s account of organizational ambidexterity examines the ability to pursue both. It does not supply a universal budget ratio. A service team may explore a new delivery arrangement while improving the reliability of its current service. Calling the first activity strategic and the second operational can hide how much judgement both require.

Start by identifying the uncertainty. Improving a known packing process asks whether an adjustment performs better within an accepted model. Exploring reusable packaging may question the model itself: who owns the container, how it returns, and whether customers will participate. Three Horizons can help discuss the portfolio, but a horizon is neither a department nor a fixed delivery date.

Separate the work and connect the decisions

Consider a hypothetical ceramics manufacturer. Its production team must dispatch intact bowls to retailers. A small group wants to explore returnable protective packaging. Both need the same experienced packing technician. Production requests help with today’s orders; the experiment needs that person to understand failure modes before designing a trial.

Workers inspect rows of ceramic bowls while another group examines different protective packing shapes.
Figure 1. Separate the work and connect the decisions.Illustration credits

An instruction to collaborate leaves the resource conflict unresolved. A workable agreement would name protected access to the technician, conditions under which production can interrupt it, and who decides when those conditions apply. The exploratory team should understand the cost it imposes on current work. Production should know which learning question justifies that cost.

Record different commitments:

WorkImmediate responsibilityUseful review question
Current packagingDeliver the agreed service reliablyWhere do breakage and rework occur?
Returnable conceptInvestigate collection and protection assumptionsWhat evidence would make a larger trial reasonable?
Shared capabilityMake packing expertise available to bothWhich commitment is being displaced?
IntegrationPrepare a possible operating ownerWhat would need to change if the trial continued?

This table is a planning aid for the fictional situation. It is not an ambidexterity score or a validated diagnostic instrument. Use assumption mapping to make the exploratory questions specific before allocating scarce expertise.

Choose an arrangement that fits the conflict

Structural ambidexterity gives different activities distinct organizational arrangements, with integration at an appropriate leadership level. In the ceramics example, that could mean a packaging exploration group with a bounded budget and a named production sponsor. Separation may protect its work, but it also creates a boundary across which knowledge and responsibility must travel.

Contextual ambidexterity puts more of the balancing work into everyday decisions within a unit. Gibson and Birkinshaw’s account of building ambidexterity describes it as complementary to structural separation. For the fictional manufacturer, a packing team might have explicit permission to run bounded improvements alongside normal work. Permission needs capacity and decision rights; an invitation to experiment during spare time is a weak operating arrangement.

Kotter’s dual operating system proposes a network working alongside a hierarchy. Its emphasis is strategic movement through a connected network. It should not be treated as another name for every exploration/exploitation arrangement. A volunteer network, a separate venture and a team alternating between tasks create different obligations.

Ask which actual conflict the proposed structure addresses. If a project needs unfamiliar partners and a different cost model, more separation may be useful. If learning depends on daily service expertise, isolation may be especially costly. A small organization may lack enough people to sustain separate units and instead need explicit scheduling and escalation.

Give leaders an integration job

A senior sponsor should resolve trade-offs that neither team can settle alone. That includes shared specialists, access to customers, use of operational systems and the conditions for a larger commitment. Escalation should be possible before a delivery failure turns the disagreement into a personal conflict.

Do not review an early exploration as though it were an established product line. Ask about the assumptions tested, the quality of the observations and the next decision. Equally, avoid protecting every experiment indefinitely under the label of learning. A question can be answered unfavourably. Innovation budgeting explains how to tie continued funding to a new, bounded commitment.

Keep the operating consequences visible. Galbraith’s Star Model brings strategy, structure, processes, rewards and people into the same design discussion. Here that means checking whether production managers are assessed only on short-term output while being asked to contribute scarce time to exploration. The framework identifies a design issue; it does not prove which reward arrangement will work locally.

Plan the transfer before the experiment grows

Suppose the packaging group wants to move from a bench prototype to a retailer trial. Before authorizing it, identify who would receive returned containers, inspect them, replace worn inserts and handle a missing return. These are proposed activities, not observed results.

Two colleagues examine a bowl and protective inserts beside an open wooden delivery crate.
Figure 2. Plan the transfer before the experiment grows.Illustration credits

Invite a possible operational owner to rehearse one complete cycle. Make room for that person to reject an unworkable assumption. A technical demonstration can establish that an insert protects a bowl under particular conditions; it does not establish that retailers will return it or that the factory can clean it economically.

The Oslo Manual distinguishes an innovation from an unimplemented idea through introduction or use. For this project, the transfer question is therefore substantive: what exactly will enter use, and who will support it? A successful exploration need not become a separate business. Its useful outcome might be a revised specification, a partnership decision or evidence that the proposed arrangement should stop.

Failure modes and evidence limits

One failure mode is a showcase unit with no route into the operating business. Another is an exploration programme that repeatedly loses its people whenever current demand rises. A third is a portfolio in which every project is called exploratory, making ordinary delivery responsibilities disappear. These are useful patterns to investigate, not diagnoses to assign from a workshop conversation.

The research contains different definitions, measures and organizational settings. A study associating ambidexterity with performance does not establish that your proposed structure causes better results. The hypothetical ceramics example demonstrates a way to reason about commitments; it is not a documented company case.

For the next review, write down the two kinds of work, their shared dependency, the person who resolves conflicts, and the condition for transferring responsibility. If those fields are unclear, commissioning a new organization chart is premature. Use an organization design inquiry to investigate the missing decisions with the people who would carry them out.

Sources: