The organizational requirements for generative AI range from low to high, depending on the use case.
Choose from the menu to explore examples
Technical pathway:
Train a foundation model from scratch
Accelerating the pace at which research scientists can identify relevant cell features for drug discovery
Accelerating the pace
High
Low
On top of the above, when training on external data, thorough legal review is needed to prevent IP issues
Process adjustments
FMs can be trained on large publicly available data, although long-term differentiation comes from adding own labeled or unlabeled data (which is easier to collect)
Proprietary data
Requires large data science and engineering team with PhD-level knowledge of subject matter, best practice MLOps, data- and infrastructure management skills
Talent
Running costs for model maintenance and cloud compute similar to above
Initial costs ~10-20X more than building on API due to upfront human capital and tech infrastructure costs.
Costs
Changing the work
Accelerating the pace
Helping relationship managers
Freeing up customer support
McKinsey & Company
Process adjustments
Processes for triaging and escalating issues to humans are needed, as well as periodic assessments of model safety
Proprietary data
A proprietary, labeled data set is required to fine-tune the model, although in some cases it can be relatively small
Talent
Experienced data science and engineering team with MLOps knowledge and resources to check or create labeled data needed
Costs
Initial costs ~2x more than building on API due to increased human capital costs required for data cleaning and labeling and model fine-tuning
Higher running costs for model maintenance and cloud computing
Freeing up customer support representatives’ time for higher- value activities
Technical pathway:
Fine-tune open-source model in-house
Freeing up customer support
Process adjustments
Processes may be needed to enable storage of prompts and results, and guardrails may be needed to limit usage for risk or cost reasons
Proprietary data
Because the model is used as is, no proprietary data is needed
Talent
Software development, product management, and database integration capabilities are needed, which require at least 1 data scientist, machine learning engineer, data engineer, designer, and front-end developer
Costs
Up-front investment is needed to develop the user interface, integrate the solution, and build postprocessing layers
Running costs for API usage and software maintenance
Helping relationship managers keep up with the pace of public information and data
Technical pathway:
Build software layers on model API
Helping relationship managers
Process adjustments
Processes largely remain the same, but workers should systematically check model results for accuracy and appropriateness
Proprietary data
Because the model is used as is, no proprietary data is needed
Talent
Little technical talent is needed—potentially for selecting the right solution and light integration work
Costs
Many SaaS tools offer fixed-fee subscriptions of $10 to $30 per user per month; some products have usage-based pricing
Changing the work of software engineering
Technical pathway:
Use software-as-a-service (SaaS) tool
Changing the work