The extensive literature on research co-creation is mostly based on problems being treated as clearly defined and solvable. What is the impact on co-creation when problems are complex, with the following characteristics: difficult to delimit, contested definitions, multiple uncertainties and unresolvable unknowns, constraints on what can be done, and no perfect solution? Co-creation on such problems requires a research mindset that appreciates that stakeholders have an essential role in making complexity evident and that limiting stakeholder involvement also limits the ability to understand and effectively act on complex problems. The Integration and Implementation Sciences (i2S) framework is introduced as a way to systematically take into account multiple stakeholders with multiple relevant inputs. In addition, one specific tool, the research-relevant modified International Association for Public Participation (IAP2) spectrum, is described as a way of exploring stakeholder engagement. For complex problems this modified spectrum can be used to describe different ways of including stakeholders, to reflect on appropriate forms of engagement, as well as to take into account both level of researcher control and different categories of stakeholder groups. A research programme examining the feasibility of prescribing pharmaceutical heroin as a treatment for heroin dependence stimulated thinking about co-creation on complex problems and is used as a practice example to illustrate the points covered.