Choosing models and tools
Using agents · topic using-agents/choosing-tools
Not every task wants the biggest model or an agent. This topic covers matching a model to a task, weighing cost and speed against quality, choosing between a chat, an agent and an unattended automation, deciding which steps stay human, and re-applying that judgment as tools grow more capable.
Concepts
- Model fit
- Matching a model to a task by what the task needs: a small fast model for classification or reformatting, a larger one for multi-step reasoning or ambiguous instructions, a model with tools where fresh facts matter. Fit is tested, not assumed: run a few representative inputs on each candidate and compare, rather than defaulting to the most capable and expensive option. glossary
- Cost and speed
- The price and delay of a choice, per task and at the volume you expect. A model that costs ten times more and answers slower must be enough better to justify it; an agent that takes twenty tool calls costs twenty times a single answer. For repeated work, small differences per call become the whole budget, so measure before scaling. glossary
- Chat vs agent vs automation
- Three ways to apply a model. Chat: you steer every turn, best for thinking and one-off drafts. Agent: you brief it and it works through steps with tools, checking in as agreed, best for bounded tasks with clear done-criteria. Automation: it runs unattended on a trigger, best for repetitive, low-stakes, well-tested tasks. Autonomy rises across the three, so the checks must rise too. glossary
- What to keep human
- The steps that stay with a person on purpose: decisions with legal or ethical weight, judgments about people, anything irreversible or costly, and the final accountability for a result. Keeping a step human is a design choice, and naming those steps early prevents them from being automated by default as tools improve. Keeping a human path also means the process keeps working when the tool is unavailable. glossary
- Reasoning across tool levels
- Tools rise in abstraction, from autocomplete to chat to agents to systems of agents, and each level hides more of what happens below. The judgment about fit, cost, risk and what to keep human has to be re-applied at each new level rather than carried over. What was safe to trust at one level may not be at the next, and vice versa. glossary
Links
- Builds on: Delegating to an agent, Capabilities and limits
- Leads to: nothing yet
- Competencies drawing on it: Chooses the right tool and autonomy level for a job
Lessons
- Chat, agent or automation (explanation)
- Reasoning across tool levels (explanation)
- The steps that stay human (explanation)
- Trying two models on the same task (tutorial)
Your reference
Each lesson above adds its takeaways and its example here once you finish it. Your reference lists every lesson you have finished.
Chat, agent or automation
Unlocks when you finish Chat, agent or automation.
Takeaways
- Chat, agent and automation are three ways to apply one model, and the difference is how many steps it does before you look.
- Autonomy rises across the three. The checks rise with it, from your reading in chat to the brief's stop points for an agent to a tested sample and undo for automation.
- The level is picked for the task's worst case. A worst case that is expensive and can't be told from the routine case keeps the task one level down.
- Each pick gets a moving rule: the evidence that moves the task up, and the new worst case that moves it back.
- The written rule stops the two silent moves, the approval that becomes a click and the task that stays in chat because it always has.
Example
This lesson has no runnable example or prompt block.
Reasoning across tool levels
Unlocks when you finish Reasoning across tool levels.
Takeaways
- Tools sit on a ladder, from autocomplete to chat to agents to systems of agents, and each rung hides more of what happens below it.
- The checks you did at the lower rung were part of the work, and they stop when the tool takes the work over. The questions about fit, cost and risk move to the level above, and you ask them there. A step you kept human needs a new place at that level, or it is gone.
- The higher-level check is only as good as the lower level under it, so a signal that doesn't add up (a round number, a fast run, a check that never fails) is the cue to go down one level and look once.
- A more capable tool makes different mistakes, and a check built for the old mistakes passes without telling you anything. Find the new check by asking what this tool gets wrong.
Example
This lesson has no runnable example or prompt block.
The steps that stay human
Unlocks when you finish The steps that stay human.
Takeaways
- An agent could take part in almost every step of an everyday process, and "could do" is a different question from "should do".
- A step stays with a person when someone has to answer for it. It also stays when the information it needs is in no system the agent can reach. And it stays when doing it keeps the person able to judge the agent's work.
- The reasons are about the person and the process. A better agent changes the draft the person starts from and leaves the list of human steps as it was.
- A step nobody named is automated by default, one busy week at a time, so the list is written when the process is set up.
- Named human steps keep the process running when the tool is down, because someone still knows how to do them.
- The boundary that holds is a named step with a named owner, and "be careful" names no step.
Example
This lesson has no runnable example or prompt block.
Trying two models on the same task
Unlocks when you finish Trying two models on the same task.
Takeaways
- Say what a task needs before you name a model: reformatting, classification, multi-step reasoning, or fresh facts.
- Find the model fit by running a few representative inputs on each candidate. The size of a model doesn't tell you its fit on your task.
- A small model is often good enough for reformatting and classification, and a large one pays off on an ambiguous request. Fresh facts need a model with a search tool, whatever its size.
- Cost and speed are per call. At the volume of a repeated task they become the whole budget, so measure before you schedule.
- Leaving the picker on the largest model is a choice too, and the table is how you check it.
Example
Turn these lines, in the order given, into a table with the columns Room, Floor, Seats, and Screen.
Harbor room, second floor, seats 8, has a screen Meadow room, ground floor, seats 4, no screen Summit room, third floor, seats 12, has a screen Quarry room, ground floor, seats 6, no screen
Sources
AEC-12Getting started with Claude Code, Agent Engineer Course (course)Brilliant ABSReasoning across levels of abstraction, Brilliant, Coding with AI skills map (reference)Academy ai-fluency-for-buildersAI Fluency for builders, Claude Academy (course)Academy building-effective-human-agent-teamsBuilding effective human-agent teams, Claude Academy (course)