There are a few things to consider when piloting or implementing Cogniti at your institution.
Please note that Cogniti is still an early-stage pilot being developed by a team at the University of Sydney, and we welcome collaborators from other institutions in our journey.
Governance #
It’s important to ensure that you follow all relevant governance in place at your institution. For example, this may include checking with your IT department, or privacy office, or educational technology group.
Our Privacy Policy and Terms of Service are available on this site.
Cogniti currently uses Microsoft Azure services for hosting the application, database, document embeddings, and GPT large language models. Cogniti can access the OpenAI-hosted GPT models, and these can be switched on/off for your organisation.
Cost #
The main costs associated with running Cogniti are server costs and AI costs. There are some economies of scale if more institutions use Cogniti’s multi-tenant solution.
AI costs, however, linearly scale with use. Every prompt and every completion costs money, and you can check out indicative costs of the Microsoft OpenAI Service.
Adoption models #
There are a few considerations here, and the best approach for your institution will depend on a balance of the following factors. All of these are to be read in conjunction with our Privacy Policy and Terms of Service.
Privacy – your institution may prefer to fully own and control the underlying services (e.g. database, document embeddings, language models), or only some of these, or none of these.
Cost – because AI (large language model) costs scale linearly with use, we cannot pay for ongoing, scaled-up use of the AI in the long term. Our collaborating institutions will need to bear the AI costs for any large-scale deployments.
Data sovereignty – your institution or jurisdiction may stipulate where data needs to be hosted. Currently, the database used by Cogniti is hosted in an Australia East region.
These considerations lead to some adoption models.
Pilot (possibly extending into early 2025): We host, and provide AI access #
For small-scale pilots, it may be beneficial in terms of ease and efficiency, balancing risk, for us to host Cogniti and sponsor some light use of the AI. In 2024, we have secured some internal funding to support other institutions in these pilots. We are working on securing funding for early 2025 to continue to provide this pilot service free of charge.
From 2025: Tier 1 (self host) #
The University of Sydney can share the Cogniti codebase with your institution, under a licence agreement, so that your institution can deploy it on its own Azure infrastructure. In this case, your institution pays all costs (e.g. servers, AI inference, support, etc). The licence agreement stipulates that if your institution makes improvements to the codebase, that you will share it back so that all Cogniti institutions can benefit.
From 2025: Tier 2 (self host, partner supported) #
We are working with a partner to enable your institution to deploy Cogniti into your own Azure tenancy. This is likely to be via the Azure Marketplace, and will attract a licencing fee plus partner charges. Your institution will also pay the costs of running Cogniti and AI costs.
While we are in early-stage pilot, we will continue to explore the cost model for this.
From 2025: Tier 3 (Cogniti Cloud, SaaS) #
We are working with a partner to enable your institution to have Cogniti hosted for you in a multi-tenant environment (similar to many other cloud software offerings available today). Your institution will be charged a fee for this service.
While we are in early-stage pilot, we will continue to explore the cost model for this.
LMS integration #
LMS integration can be arranged on a per-organisation basis. Your institution’s LMS administrator will need to reach out to us and we can provide configuration guides.
Cogniti currently supports LTI 1.3 integration with Canvas and Moodle. Integration with Blackboard is under development. Integration with other LMS platforms can be discussed.
Cogniti supports API integration with Canvas.