Digital Innovation Agents

Appearance

Ideation methods ​

Once you have a problem worth solving, you need ideas that try to solve it. Ideation is the part where the team goes wide, makes a mess, and then narrows down on the candidates worth prototyping. The Business Analysis skill will propose the right ideation method depending on where you are stuck: empty page, too many weak ideas, or one strong idea that is not yet concrete enough to test.

Like the discovery methods, these are run between humans. The agent waits for you to come back with a shortlist of ideas it can turn into an Idea Canvas or a Value Proposition.

Index ​

Brainstorming ​

The classic group ideation session. People in one room, one question, short timebox, post-its on a wall. It produces volume, not quality. Quality comes later in the selection step.

When to reach for it. You have a sharp How-Might-We question and an empty solution space. You need a wide spread of candidates before narrowing down. The team has not yet tried to solve the problem together and you want the fast-thinkers and the slow-thinkers to hear each other.

How to run it.

  1. Write the question somewhere everyone can see. One question per session. Never two.
  2. Brief the eleven rules aloud before starting: go for quantity, go for wild ideas, build on ideas of others, defer judgement, one voice at a time, stay on topic, fail early and often, take fun seriously, work visually, work multidisciplinary, think from the user's side.
  3. Set a timer for 15 to 20 minutes. One idea per post-it. Everyone writes in silence for the first five minutes, then starts sticking them up.
  4. Encourage weird variants. "What if the constraint was removed?" "What if the opposite were true?"
  5. Cluster the post-its at the end but do not evaluate yet. Evaluation is a separate session.

Team and time. Four to six people is the sweet spot. 15 to 20 minutes of writing, 10 to 15 minutes of clustering.

Things that go wrong.

  • Critique sneaking in. One "yes but" from a senior team member and the room goes silent. The moderator has to shut it down immediately.
  • Letting fast talkers dominate. The silent five-minute opener prevents this if you enforce it.
  • Running the session with people who all have the same background. The wildest ideas come from someone whose daily job is nothing like yours.
  • Forgetting to timebox. Brainstorming should feel slightly too short.

What to bring back to the agent. The 8 to 12 most interesting candidates from the clustering, as short phrases. The agent will help sharpen them into Idea Canvas entries.

Brainwriting ​

Silent ideation. Everyone writes in parallel, passes their sheet, and builds on what the next person wrote. It avoids all the social dynamics of brainstorming and usually produces a broader range of variants.

When to reach for it. Previous brainstorming sessions were dominated by one or two voices. The team is introverted or multicultural and the group rhythm of brainstorming does not work. You already have a few seed ideas and want to see how far they can travel.

How to run it.

  1. Give each person a sheet with three columns and six rows.
  2. Set a three-minute timer. Everyone writes one idea in each of the three columns of row one.
  3. When the timer ends, pass the sheet to the next person.
  4. They read the three ideas above and build three new variants or counter-ideas in row two.
  5. Repeat until the sheet is full or the energy drops.
  6. Read the sheets aloud at the end and cluster on the wall.

Team and time. Six people is ideal, the classic format is called 6-3-5. 30 minutes total.

Things that go wrong.

  • Letting the round drag. If anyone is still writing after three minutes, pass anyway. Under-pressure is better than over-pressure for this method.
  • Building on someone else's idea without reading what they wrote. The whole point is the chain.
  • Turning brainwriting into brainstorming halfway through because "it is more fun this way." The silence is where the value lives.

What to bring back to the agent. All the ideas from all sheets, roughly clustered, with the thread you found most promising highlighted.

Idea tower ​

A technique for iteratively enriching a single idea instead of generating new ones. People take turns adding one element, one constraint, or one twist. The result is usually a concept far richer than the seed.

When to reach for it. You already have a seed idea that everyone agrees is promising, but it is still too thin to prototype. The team keeps saying "yes but we would also need X" and never closes the loop. You want to push one concept to its limit before comparing it to alternatives.

How to run it.

  1. Write the seed idea at the base of a flipchart. One sentence.
  2. Each person takes one minute to add one new element to the idea. Features, users, constraints, twists, environments. Additive only.
  3. Keep going in rounds until the energy drops or the idea becomes coherent enough to draw.
  4. Then sketch the final version on a second sheet.

Team and time. Three to six people. 20 to 30 minutes.

Things that go wrong.

  • Critiquing instead of adding. The rule is strict: every contribution adds, nothing removes.
  • Taking too long per turn. One minute is enough to force commitment.
  • Stopping too early because the idea already "feels good." The tower only gets interesting after round four.

What to bring back to the agent. The enriched idea as a sketch plus a short list of the most surprising additions that came up.

Collective notebook ​

A slow ideation method. Each team member keeps a personal notebook for a week or two, captures ideas when they surface naturally, and shares at a fixed interval.

When to reach for it. The problem is complex and the team needs time to let it sit. One-hour workshops keep producing the same shallow ideas. You want ideas from moments outside work: in the shower, on the commute, while cooking.

How to run it.

  1. Give every participant a notebook (paper or digital) and a clear prompt.
  2. Set a duration: one or two weeks works best.
  3. Ask participants to log any idea the moment it appears, even if it feels silly.
  4. At a fixed interval (every few days), everyone writes a short summary of their best three or four entries and shares it in a shared doc.
  5. At the end of the period, run a single synthesis session to cluster and pick the strongest candidates.

Team and time. Three to eight people. One to two weeks of elapsed time with light daily effort.

Things that go wrong.

  • Forgetting to check in. Without the mid-period share, people quit after the first day.
  • Over-structuring the notebook template. Leave it loose. A tight template turns the notebook into a chore.
  • Trying to evaluate ideas during the period. Keep all evaluation for the final session.

What to bring back to the agent. The strongest ideas from the synthesis session, with a short note on which context or moment produced each one.

Inspiration cards ​

A divergent-thinking technique using random stimuli (cards, images, words, analogies from other domains) to unstick the team when every obvious idea has already been said.

When to reach for it. The team is repeating variations of the same three or four ideas. You want to force unfamiliar associations. You need ideas that could not have come from your direct industry experience.

How to run it.

  1. Prepare a small deck of cards. Words from a random domain, images from unrelated contexts, well-known products from other industries. 20 to 40 cards is enough.
  2. Shuffle and draw three or four cards at random.
  3. For each card, the group has two minutes to generate ideas that would connect the card to your How-Might-We question.
  4. Capture every idea, including silly ones.
  5. After the round, cluster and pick the strongest surprises.

Team and time. Three to six people. 30 to 45 minutes per round.

Things that go wrong.

  • Reaching for the "safe" card interpretation. The fun comes from the weird one.
  • Picking a deck that is too close to your domain. If all the cards are software products, you will only get software ideas.
  • Killing the round after one card. The good cards are the third and fourth.

What to bring back to the agent. The surprising associations and the ideas that came from them.

Idea clustering and selection ​

A structured way to go from "too many ideas" to "a shortlist worth testing." Ideas get clustered by theme, then scored against a small number of criteria so the team can defend the selection later.

When to reach for it. You have 30 to 50 ideas on the wall and no idea which one to build first. Different team members advocate different candidates based on personal preference. The BA is asking you for the top three ideas and you cannot pick.

How to run it.

  1. Cluster the post-its by theme (not by quality yet). Name each cluster.
  2. Pick a small set of criteria. User value, feasibility, transferability, or whatever fits your project. Four is usually enough.
  3. Score each cluster or each leading idea against each criterion on a 0 to 5 scale.
  4. Total the scores. The top cluster is your first shortlist candidate.
  5. Apply a sanity check. "If we only built this one, would we be proud?" If not, go back to the criteria.

Team and time. The original ideation team. 60 to 90 minutes.

Things that go wrong.

  • Picking too many criteria. With more than four, the scores become noise.
  • Letting the highest number win automatically. The scores are an aid, not a decision.
  • Clustering by quality instead of theme. You end up with a "good ideas" and a "bad ideas" pile, which hides structure.

What to bring back to the agent. The top three ideas with their scores and a short paragraph on why they made the shortlist.

Jobs to be done ​

A framing method from market research and innovation theory. It asks "what job does the user hire our product to do" at three layers: the functional job, the emotional job, and the social job. It is less about generating ideas and more about understanding why an idea would win.

When to reach for it. You have an idea but cannot explain why users would actually switch to it. You keep writing user stories that only cover the functional side and feel hollow. You want to sharpen the value proposition before you commit to prototyping.

How to run it.

  1. Write the idea at the top of a sheet.
  2. For the target user, answer three questions:
    • What functional job are they hiring this product to do?
    • How do they want to feel while doing it?
    • How do they want to be perceived by others while doing it?
  3. For each layer, write the hiring criteria (reasons to switch to the new product) and the firing criteria (reasons to abandon it).
  4. Compare the layers. The deepest, most specific layer is usually where the real value lives.

Team and time. Two to four people. 60 to 90 minutes.

Things that go wrong.

  • Only filling in the functional layer. Emotional and social are where the stickiness lives.
  • Writing the jobs from the team's perspective instead of the user's. Use direct quotes from interviews where possible.
  • Stopping after hiring criteria. The firing criteria tell you what kills adoption.

What to bring back to the agent. The three jobs with hiring and firing criteria. The agent uses these directly when writing user stories in Requirements Engineering.

TRIZ ​

A systematic inventive-problem-solving method from Soviet engineering. You express the problem as a contradiction (two requirements that cannot both be satisfied at once), then use a catalog of inventive principles to resolve it.

When to reach for it. The problem has a genuine technical contradiction. "We need X and the opposite of X at the same time." Conventional brainstorming produces only incremental variants. The domain is engineering-heavy and precedents from other fields might help.

How to run it.

  1. State the contradiction in one sentence. "The product must be strong and light." "The process must be fast and safe."
  2. Look up the inventive principles that apply to that contradiction in the TRIZ contradiction matrix.
  3. For each relevant principle, generate one idea that applies it to your problem.
  4. Evaluate the ideas and pick the most promising two or three to prototype.

Team and time. Two or three people, at least one with some TRIZ background. 90 minutes to half a day.

Things that go wrong.

  • Trying TRIZ on a non-technical problem. It can work, but the contradiction matrix is tuned for engineering.
  • Using the principles as buzzwords instead of applying them concretely.
  • Skipping the contradiction step. Without a precise contradiction, TRIZ has nothing to work with.

What to bring back to the agent. The contradiction, the principle you used, and the resulting idea.

Kill your company ​

An adversarial exercise. The team pretends to be a startup attacking your own company, looking for the weakest points and the fastest ways to disrupt you. It surfaces assumptions and strategic risks that rarely come up in ordinary ideation.

When to reach for it. You suspect your own team is too close to the current product to see its weaknesses. Competitors are moving fast and you cannot articulate where the threat lies. You want to pressure-test the value proposition against an imagined adversary.

How to run it.

  1. Flip the frame. "We are now a startup. Our mission is to destroy this company."
  2. Brainstorm attack strategies. Weakest customer segments, easiest feature to copy, cheapest price point, boldest brand move.
  3. Rank the attacks by plausibility and damage.
  4. For each top attack, write the defence. What would the current company have to do to prevent it?
  5. Identify which defences are already in place and which are missing. The missing ones become work items.

Team and time. Four to eight people, ideally with at least one outsider. 90 minutes.

Things that go wrong.

  • Pulling punches. If the team is too loyal to the current product, the attacks stay toothless. Invite someone who does not owe anyone anything.
  • Stopping at the attack step. The value is in the defence work, not the attack.
  • Treating the output as criticism of the current team instead of intelligence about the future.

What to bring back to the agent. The top attacks and the missing defences, in priority order.

Next steps ​

  • Once you have a shortlist of ideas, the BA agent will ask you to turn them into Idea Canvas entries and sketch a Value Proposition.
  • For testing the resulting ideas against real users, continue to Validation methods.
  • For the discovery methods that produced the input to this phase, see Discovery methods.